Blunder Valley

Not Dead Yet: The Steelport Data Centre Fight Continues

2026-06-09T00:00:00Z

image credit: R. Wignall

Note: For vital background and context on the Steelport data centre(s), see the previous Blunder Valley post.

A crucial first step

On June 4th, Hamilton's Committee of Adjustment voted to deny a land severance application (pdf) by Slate Asset Management which, if approved, would have accelerated Slate's plans to build a hyperscale data centre by the waterfront.

The denial was in no small part due to the overwhelming response from Hamiltonians; the Committee member who moved to deny the application even cited the collective effort as the reason for the denial. Residents showed up in large numbers - thousands signed petitions, nearly 1800 sent written submissions to the Committee, and dozens spoke at the meeting itself, for nearly 8 hours in total.¹

Perhaps most importantly, around 1000 people showed up physically outside City Hall at 8:30am on a Thursday - on just a few days' notice - making it absolutely clear that we will not allow those with money and power to unilaterally impose these data centres on our city.

But despite this win, the fight is far from over.

Slate will appeal this decision

This victory only represents a temporary reprieve. As of right now, there are still multiple viable paths for Slate to build a hyperscale data centre on the old Stelco lands on the waterfront.

First: Slate is likely to appeal the decision to the Ontario Land Tribunal (OTL). We will find out soon if that's the case; the appeal can be filed at any time within 20 days of the decision (see Ontario Planning Act s.45(10) and s.45(12)).

After that, there's no definite timeline on how long the appeals process could take. Some estimates seem to be in the range of a month or two. Others reckon the wait might be much longer. I'm not counting on the latter. When people in power really want things to move quickly, they usually have ways of greasing the wheels.

Worse: there's a good chance Slate would win an appeal. For one thing, the Committee of Adjustment's verbal explanation of their decision amounted to saying 'the application technically checked all the boxes, but just look at this crowd'. Plus, well, this is Doug Ford's Ontario we're talking about here, where the rules are often worth about as much as the paper they're printed on.²

In other words, this is all far from settled.

If Slate wins their appeal at the OTL: going forward, the severed parcel would only be subject to further scrutiny/approvals from city staff. Meaning, no more official means of input from the public, or from our elected representatives on Council. There are still a list of 16 conditions attached to the land severance, but many of these simply require Slate to submit reports and assurances on various issues. Beyond those conditions, staff would presumably just follow the existing zoning bylaws which, as Slate never gets tired of repeating, currently allow for hyperscale data centres.

Next Steps

First:

Hamilton's Ward 3 City Councillor Nrinder Nann has put forward a motion asking City staff to report back on factors that might inform a Municipal Framework around data centres in Hamilton. That report will be coming back on June 16th. We don't yet know exactly what staff will say, or what such a Framework might ultimately entail (size limits? a data centre moratorium? strict regulation around noise, water, heat?), but it is important to contact councillors before June 16th to share your concerns.

Implementing a Framework would, at the very least, create hurdles that data centre projects would need to clear before being approved. And in the best case scenario, it could mean a complete moratorium on data centre construction in this city.

Even that best case, however, wouldn't necessarily be an absolute victory for Hamilton residents - it would merely force higher levels of government to resort to more draconian tactics and override City Council, if they wanted to get these things built.

Second:

Municipal elections are coming at the end of October. We will need clarity from candidates running for Mayor and Council. StopTheDataCentre.ca has made a tracker that shows candidates' positions. We'll be keeping an eye on this, especially as the election nears. Who will support residents of Hamilton in this fight - and who will support the top-down imposition of these massive AI data centres?

(by the way - if you're in another municipality and want to make a tracker of your own: get in touch and we can get you connected with the right folks!)

That's the basic rundown on what's been happening. If you're desperate for some more juicy tidbits, here's...

Some More Juicy Tidbits

Lobbying

One thing that wasn't covered in the previous Steelport post is Slate Asset Management's lobbying of the federal government. Slate has spoken to multiple federal entities, including:

the PMO
Finance Canada
Innovation, Science and Economic Development (ISED)
Housing, Infrastructure and Communities (HICC)

And each of their lobbying activities lists the same discussion topic:

"Engage government decision-makers regarding Canada’s Sovereign AI Compute Strategy, including advancing Slate’s upcoming proposal to the Digital Research Alliance of Canada [DRAC], and opportunities to advance digital innovation and clean growth."

The word "including" is doing a lot of work here - it's now reasonable to assume they've been discussing other data centre plans, beyond just the DRAC project. This kind of phrasing reflects the sneaky way Slate has been trying to hoodwink well-meaning community organizations (such as Environment Hamilton, among others) by pitching them a nice little cute smol bean research facility - all while a hulking hyperscale monstrosity looms in the background. And to be clear: we don't actually have any actual specifics on the DRAC data centre!³

The PMO

It's not entirely clear why Slate is lobbying the Prime Minister's office directly. We do have indications that many decisions are coming from the top, however. Possibly of note: one of three investors in Slate's €425 million ($685m) debt refinancing round in 2024, a bank called Natixis, opened a Toronto office inside Brookfield Place in 2023, and in 2022 underwrote Brookfield's DCI Data Centers portfolio.

Finance Canada

There are question marks surrounding this lobbying as well. My best guess: it has something to do with the disbursement of whatever funding might come out of the other lobbying activities discussed below.

ISED

The reason for this lobbying seems more obvious: ISED is the department responsible for doling out funding related to the federal government's AI Sovereign Compute Infrastructure Program (AI SCIP).

Slate has partnered with the Digital Research Alliance of Canada (DRAC) and Hamilton Community Enterprises (HCE) on a proposal, submitted to ISED, to construct a "research" AI data centre just south of the land where Slate is still hoping to build a hyperscale facility. (See the earlier Steelport post on Blunder Valley for background on this.)

And while a federally-funded "sovereign AI research" facility may sound good on paper, there are serious questions that we should be asking.

For example: it was recently revealed that the federal government has been secretly signing contracts with US-based data analytics firm Palantir - a company closely aligned with the Trump administration, and who have been trying to become essential infrastructure for the UK's National Health System. Meanwhile, Canada's "AI champion", Cohere, is scrambling to distance itself from its past reported connections to Palantir.

To be clear: there's no direct evidence to suggest that companies like Palantir will get access to Canadians' sensitive personal data. But the links between the company and various levels of government appear quite strong: up until April 2026, David MacNaughton, former President of Palantir Canada, who in 2020 was found guilty of violating Canada's ethics laws for undisclosed lobbying on behalf of Palantir, was a member of the Prime Minister's Council for Canada-U.S. Relations.

HICC

This one is a little more ambiguous; there are two likely reasons that Slate would want to lobby HICC regarding "digital infrastructure". Both relate to the fact that the Canada Infrastructure Bank (CIB) reports to HICC.

First possibility: in Budget 2025, the federal government made changes that allow the CIB to invest in AI data centres. Slate may be seeking financing from the CIB to realize their hyperscale AI data centre ambitions.

Second possibility: Slate could be lobbying the HICC for funding to build out Hamilton's district heating system to use waste heat from either the DRAC data centre, or a hyperscale facility. This fits with their partnership with Hamilton Community Enterprises, who in 2024 commissioned an Energy Harvesting Study to examine the idea of pumping heat from Hamilton's industrial sector via hot water pipes into buildings in the downtown core.

District Heating and other "Green" initiatives

There's nothing wrong with the idea of a district heating system, on its face. Capturing waste heat and re-using it is fantastic in principle. But the devil, as always, is in the details of implementation.

For one thing, the Energy Harvesting Study based its emissions-reduction claims on a comparison with fossil-fuel heating. Which makes the numbers look great! But these days we have many other economical low-emission heating methods, including air-to-air heat pumps and geothermal heat pumps.

Compared to resistance heating (e.g. an electric baseboard heater), these heat pump systems are 300% to 400% efficient - meaning that per unit of electricity used, they heat a building 3x to 4x as much as resistance heating does.

Computers are, insofar as they generate heat, resistance heaters. So here's what we should be asking: is this inefficient heat source, also known as an AI data centre, creating enough benefit - and causing a low enough level of harm - to justify the massive infrastructure expense involved in such a system? Especially compared to simply going geothermal, which uses a fraction of the energy? Do we want to run headlong toward inextricably linking our downtown buildings' heating systems to an industry that isn't providing much social or economic benefit?

Do these AI data centres create many jobs? No. Do they have a long list of environmental impacts? Yes. Have they caused electricity prices to rise everywhere they get built? Yes. Are AI data centres a financial bubble that could pop any moment and leave our downtown buildings scrambling to find a new heat source? It seems risky!

Besides: where does all this waste heat go in the summer, when those same buildings need cooling? A geothermal system can do cooling and heating; a waste-heat-capture system does not.

(For more on this topic, see Dr. Anne Pasek's excellent talk titled "Data Centres in Ruins: What Comes After the AI Bubble Pops?").

Like a lot of other development - the AI data centre bubble is the perfect example of this - there's a bias toward "innovation" without ever considering if that innovation truly represents progress. You also have to wonder if HCE, in this instance, is like a hammer looking for a nail, in their quest to expand the district heating system.

HCE's Annual General Meeting is on June 23, 2026. I'm hoping that some City Council members will start asking some difficult questions about HCE's partnership with Slate Asset Management, and take a closer look at both the benefits and the drawbacks of initiatives like these.

Thanks for reading. If you enjoyed this piece, please share it with friends and family who you think will be interested. And if you have a news tip about Steelport or any other data centre project, send me an email or find me on Signal.

Notes

To give a sense of what this Committee might ordinarily deal with: a separate agena item was requesting that "an eave or gutter ... may encroach into any required yard to a maximum of 0.5 metres instead of the maximum permitted 0.45 metres", as part of a property owner seeking permission to build a two-storey backyard housing unit and garage. In fact, there was a surreal moment early in the day when someone called in to present their verbal concerns to the Committee of Adjustment regarding this exact agenda item, clearly not knowing what the situation looked like inside City Hall. back

There is a tiny sliver of hope here: a number of the written and verbal comments submitted to the C of A seemed to make cogent and compelling arguments for denial based on facts around the land severance application itself. But I'm not a lawyer, and it feels like this is a long shot, especially in the face of a federal and provincial government that seem determined to ram through favoured projects regardless of opposition or any social & environmental consequences. back

To be clear: the "small" facility being proposed by DRAC is still likely to draw enough power for well over 10,000 homes. But we don't know the exact number. Because although the organization has loads of time to assure the CBC that everything will be low-impact, I've directly asked DRAC on more than one occasion for information about power draw, water consumption and discharge, cooling type, and specific metrics and standards for noise and heat island effect mitigation, and thus far have only received vague non-answers as a response. back

The Steelport AI data centre(s): what we know so far

2026-05-19T00:00:00Z

UPDATE, May 30:

Please visit StopTheDataCentre.ca for information on how to get involved, as well as a toolkit for engagement with the upcoming June 4th Committee of Adjustment meeting.

UPDATE 2, May 27:

Slate has submitted a 'land severance' application to Hamilton's Committee of Adjustment (see earlier update).
Members of the public can submit written comment up until 12:00 noon on Tuesday June 2, 2026 and it will be included in the June 4th meeting agenda. In-person delegations are also permitted.
The city's instructions for doing so are here.
The email address is cofa@hamilton.ca. Be sure to include your name, address, and the file number, which is B-2026-00011.

UPDATE, May 27:

there is a LeadNow petition demanding proper public consultation before allowing any data centres in Hamilton
See also Slate's severance application tabled May 19th which specifically references hyperscale data centres

Correction:

May 30: this article has been updated to reflect the fact that M5 zoning also allows for a "Manufacturing" use case, which, like the "Research and Development" use case already described, contains within its definition the "Computer, Electronic and Data Processing Establishment" use case. We regret the error.

Background

On March 2nd, Canada's National Observer reported on proposed data centres in Ontario that would total 2 gigawatts of new load on the electrical grid.

One of the reported data centre proposals is from Slate Asset Management, owners of the 800-acre former Stelco lands on Hamilton's waterfront, a site the company has branded "Steelport".

Slate's original 2025 application to Ontario's Independent Electricity System Operator (IESO) was for a 180 MW connection, and the request explicitly specified "data center" as the load type, according to my conversation with the journalists at the National Observer who broke the story.

Slate's connection request has since been updated to 400 MW, but no longer specifies a load type.

Into the weeds

There are two main issues:

The Digital Research Alliance of Canada (DRAC) has selected Steelport as the site for their bid to receive $705 million in federal funding to build an "AI supercomputer" data centre, as part of the feds' AI Sovereign Compute Infrastructure Program (AI SCIP).

If the site is selected by the government, the facility would be built by Slate, at which point the land and facility would be sold to DRAC, a non-profit funded by the federal government.

There are indications that Slate is also looking to build other (potentially much larger) data centre infrastructure at the Steelport site: not only did their original application to the IESO specify "data center" as the load type, but an April 8 presentation by Slate also suggests they are eager to build a hyperscale data centre on Hamilton's waterfront.

Regardless of what happens with either of these possible projects in the near term, my gut instinct at this point is that Hamiltonians need to push for one or both of:

capping the size of data centres in our city - I'd suggest 1-megawatt as a potential starting point - and/or a moratorium on building data centres for a specified time period, e.g. 5 years;

amendments to zoning by-laws to require special approvals for data centres above a certain power requirement.

Part 1: The DRAC data centre

We'll begin with the likely boondoggle of the DRAC data centre.

There has been very little transparency about this project's specifics. In response to a question about power consumption, Ward 4 Councillor Tammy Hwang provided me with an estimate of 8-10 MW, but noted this number was only to the best of her recollection.

Slate, although they have answered some of my questions about the project, has declined to say what the power draw of this facility would be. DRAC, meanwhile, has not yet responded to any of my questions or even acknowledged my emails, one sent on May 11th, and a follow-up on May 15th.

So, right now the best information I have is that the DRAC data centre would be an 8-10 MW facility, at a cost of $705 million.

The numbers don't add up

It's difficult to find a way to make these numbers work. According to an April 2026 piece in Canadian outlet The Logic:

"AI data centres can cost between $19.5 million to $33.5 million per megawatt, with most of that down to buying IT equipment, according to an October 2025 analysis by the federal innovation department."

Even if the DRAC project was at the highest cost in that range, that leaves $370 million to be accounted for, before adding in staffing, maintenance, and electricity.¹

I've put the rest of the math in the footnotes, but here's the upshot: if this is a 10 MW data centre, then including the 7 years of operational budget the funding includes, my best estimate is that the total cost is just over $400 million.²

That would mean around $300 million in public funding is vanishing into thin air.

Or: maybe the data centre is much larger than the 8-10 MW estimate Councillor Hwang was given. This could explain where the $705 million in public funding is going, but on the other hand, a larger data centre is even more of a concern for Hamiltonians.

The DRAC data centre building

As shown in the presentation by Slate on April 8th, 2026, the proposed location of the DRAC facility is an 8-acre building located in the block between Industrial Drive & Burlington St East, and Gage Ave & Depew St. The existing building at that location looks to be roughly 8 acres.³ Slate declined to say if they plan on retrofitting the building or starting anew.

Importantly, the location is the part of the Steelport land closest to a residential neighbourhood - just 500-600 metres north of the nearest homes. I've asked DRAC what mitigations they would require to be put in place in terms of noise, heat island effects, etc. and will update this post if/when I receive a response.

Data centres are harming local grids

Data centres' voracious water and energy consumption are well-known (and I'll return to those aspects later), but there are other effects that have gone under-reported.

A 2024 Bloomberg investigation found a strong correlation between proximity to "significant data center activity" and grid power distortions that "can force home electronics to run hot, or even cause the motors in refrigerators and air conditioners to rattle". Furthermore, the authors write, "[t]he worse power quality gets, the more the risk increases. Sudden surges or sags in electrical supplies can lead to sparks and even home fires."

The cause of this isn't the electricity consumption itself, but how rapidly data centres switch large loads on and off. And they're hammering grids with these load spikes on such a regular basis that they may be causing neighbouring households "billions of dollars in total damage" to their appliances.

The piece goes on: "More than half of the tracked households showing the worst distortions of power quality are located within 20 miles of significant data center activity" - "significant data center activity" being defined as anything over 10 MW.

So, while a 10-megawatt data centre might sound small, even quaint, in an age of major tech corporations announcing gigawatt-scale data centres,⁴ these facilities are having all sorts of negative impacts on nearby households.

Besides, as already mentioned, there's the distinct possibility that the DRAC data centre would be much larger than the 8-10 MW estimate I was given.

Other externalities

Water & heat

The Hamilton Spectator recently reported that the facility is proposing to use evaporative cooling - which raises concerns about a heat island effect in the surrounding area (pdf).

The same article imples that the water for cooling would be pumped out of Lake Ontario. But data centres typically cannot use water containing live organisms, which causes biofilm build-up on equipment; they instead require potable water (in this case, likely from the municipal drinking water system).

According to Councillor Hwang, there is also legacy water treatment infrastructure at the Steelport site, but I haven't been able to confirm that it's in working condition, and Slate's slide deck doesn't mention it either. I've asked DRAC if they're looking at entering into any long-term agreements with Slate for water supply or any other service/resource - but as previously noted, I haven't yet received any response.

Noise

In terms of noise: my understanding is that noise impacts are definitely an issue over a radius that would impact a lot of people's homes, but I haven't yet dug deeply into that particular topic. Data centres tend to run at all hours of the day and night, though - and this building wouldn't even be the only source of bothersome all-night hum affecting residents in this part of the city.

Data protection

Lastly - the supposed reason the feds would giving DRAC $705m for this project is to create "sovereign AI infrastructure", ostensibly because we don't want to be processing Canadians' sensitive data on US servers. Well, Slate Asset Management has billions of dollars invested in US real estate, a situation that is easily leveraged by the US to undermine the security of this facility. Whether or not we think the data centre is a good idea: is this the company we want in charge of building it?

But wait, this data centre is for research - That's good, right?

I asked myself the same question. And I think the answer depends on what kind of "AI" we're talking about here - because "AI" is, ultimately, a marketing term that's being used to describe a number of very different kinds of computational systems.

Machine Learning (ML), for example, has been around for decades, and there are plenty of use cases where it functions reliably and usefully. Protein folding simulations are one example. Early voice-to-text transcription software is another. A lot of ML applications can easily be run on a home PC, or a couple of computers in a company's IT department.

Recently though, we've seen an explosion in more computationally intensive systems, including Large Language Models (LLMs) like ChatGPT, Claude, and Gemini. Barely a day goes by without a new story of how unreliable - and even dangerous - these systems are. Here are two recent examples that hit close to home:

On May 13 reports emerged that an LLM, intended for use by Ontario doctors as a transcription tool, was found to be generating "incorrect, incomplete and hallucinated information", including suggesting therapies that the doctor had not even mentioned.

In March, McMaster University research fellow and guest teacher Kémy Adé had her permanent residence application rejected after an LLM, used by the government in processing the application, invented a job description that didn't match her (real) stated work experience.

In fact, large language models are causing so many problems in the scientific research community that the widely used open-access academic preprint site ArXiv.org announced on May 15 that any authors who submit papers found to contain LLM-generated errors will be banned from the site for a full year.

And yet: the federal government is quite explicit that their primary reason for wanting to build "AI compute infrastructure" is to provide LLM researchers with access to large clusters of GPUs.

It's maddening. Over a trillion dollars have been sunk into the development of LLMs, and they're still completely unreliable. In fact, many researchers argue that LLMs are inherently unreliable by their very nature.

And even if their problems were magically solved tomorrow: do we really need this huge data centre? It's been barely six months since a $42.5m U of T data centre was announced, and only seven months since the ribbon was cut on a $43m project in Waterloo - a facility that in October 2025 was hailed as being "designed specifically for immense scale." And already we need another one that's 15 times the size? For what, exactly?

Part 2: Other Data Centres at Steelport

As mentioned, Slate has so far refused to answer most of my questions, not only about the size of the DRAC data centre, but also whether they're looking to build other data centre infrastructure at Steelport.

Initially, after reading the National Observer's reporting, I asked Councillor Hwang about Slate's data centre plans; she explained the DRAC data centre, and informed me that the City has told Slate they'd "prefer they don't build an Amazon" (or other hyperscale data centre), but beyond that, what I was told is this: Slate owns the land, and what they do with it is up to them.

And the Steelport site is indeed zoned to allow data centres, seemingly of any size; the M5 zoning (pdf) allows for, among others, "Research and Development Establishment" and "Manufacturing" use cases, which, according to their definitions (pdf), allow for a "Computer, Electronic and Data Processing Establishment" - a definition last updated in 2010, surely never meant to include the massive hyperscale data centres of today.

But hey, maybe Slate will respect Hamiltonians' wishes out of the goodness of their hearts?⁵

I'm not counting on it. In Slate's April 8 presentation, there's a section clearly separate from their discussion of the DRAC project, titled "Data Centre Opportunity at Steelport". It talks up the potential of the 800-acre site, and points to "400+ MW available" electricity, with "1 GW potential". Forget alarm bells, that's air raid siren stuff.

Another slide boasts of the potential to create "30K+ new jobs". But data centres aren't exactly engines of employment. During contruction "a large facility might require more than 1000 people" - which to be fair, isn't nothing - but once built, it employs only a few dozen workers. Besides, most of the eye-popping build costs don't go into the pockets of local tradespeople - they purchase extremely expensive computer hardware made by a US company, Nvidia. ⁶

Immediately following that section is one that pitches a partnership with Hamilton Community Enterprises (HCE) to build a system to pipe waste heat from the Steelport site into buildings in the downtown core, linking it with the existing district energy system. Not a bad idea on its face - but here's the problem: even mainstream investment firms have recognized for some time that today's AI infrastructure frenzy is a massive investment bubble. Does it make sense to spend an enormous amount of city funds to build a heat corridor that won't have a heat source if/when the bubble pops? Why not install high-efficiency heat pumps, or geothermal instead? Why make our public systems dependent on the existence of a particular industry?

The point here is, Slate is clearly trying to pitch a hyperscale data centre, and is willing to tell Hamiltonians all sorts of fanciful stories about the supposed benefits. This doesn't mean they have clients lined up for one, or that they know the first thing about building them, or that this pitch is anything more than the asset manager flailing around trying to find tenants for 800 acres of lakefront property, and hopping on the 2024 AI data centre hype train.⁷

But regardless of their capability: if Slate finds an "Amazon" or "OpenAI" willing to sign a deal for AI compute, I'm certain they'll at least try to build a data centre. I also worry that HCE may quietly be on board with the idea.

Local and regional concerns

Hyperscale data centres bring a multitude of real and serious concerns. Like their smaller counterparts, there are the same issues of noise, heat island effects, and grid disruptions, except these are multiplied by an order of magnitude or more. Water consumption starts to be a major problem at these sizes.

And there's the issue of increased electricity rates for Ontarians. To produce all of this new energy for data centres, both the federal and provincial governments are pushing to build nuclear plants, which end up costing ratepayers several times per kWn what solar and wind does. [YouTube; Ontario Clean Energy Alliance, timestamped]

Increased energy use leads to increased carbon emissions: not only has Ontario increased the amount of energy generated by high-emissions gas plants over recent years, the proportion of energy from gas has also risen [YouTube, timestamped]. Huge projects like these, which offer little social or economic benefit, mean we will continue to move backwards on climate.

What's more, if (when) the AI bubble does indeed pop and these AI data centre projects go under, Ontario electricity ratepayers will still be on the hook for the cost of the generation and transmission infrastructure built to service them.

Suggested actions

Hamilton needs by-laws that meet the current moment. Large data centres have outsized impacts on residents and ratepayers, and it's unlikely those who wrote the relevant by-law definition in 2010 had any idea how massive these projects would become. So an update is essential. Maybe something like, "any data centre project must be approved by Council (or Planning Committee) vote? I'm not a lawyer or by-law expert, much less someone who's accustomed to navigating things at City Hall, so I'd be really interested to hear your thoughts.

Push for a cap on data centre size. My initial thought was to demand something similar to the 10 MW limit proposed in Maine, but given the issues around noise and grid strain, I'm wondering if this is still far too high. Now I'm thinking more like 1 MW; again, please do get in touch with your thoughts.

If you liked this post: I'd love to hear from you. Let me know which parts you feel are the most important or helpful. Please consider sharing it with any friends you think will be interested. And/or consider contacting your city councillor to let them know how you feel about this project, or AI data centres in general.

Thanks for reading!

Notes

I also attempted a high-ball estimate of the electricity cost for a 10 MW data centre: even assuming it runs at full capacity with zero downtime, and pays much higher than average industrial electricity rates around the clock, the energy expense still only works out to around $4m/year. The real number would certainly be far lower. back

To estimate the operational costs, we can use existing data: a $42.5 million AI compute facility unveiled in November 2025 at the University of Toronto is expected to spend $40 million on capital costs, and $1.25 million each year on "talent and operational costs" for a two-year period, according to DRAC's own numbers.
Even assuming staffing costs increase linearly with capital costs - which they generally don't - that's still only $10.5m/year for a $705m data centre.

And according to the Canadian Government, this funding is for 7 fiscal years of operations. $10.5m/year x 7 years = $73.5m. Add this to the $335m development costs, and it comes to $408.5m.

Secondly, regarding the discrepancy in the funding numbers between what Slate & DRAC mention vs what is on the feds' AI SCIP website: the $705 million number has apparently been increased by the federal government to $890 million; I have yet to track down an explanation for the change. Here's some reporting from late 2025 that contains the $705 million number; the description of a "large-scale, Canadian-owned supercomputer" is virtually idential to the feds' description of what they're now offering $890 million in funding for. Furthermore, both Councillor Hwang and Slate directed me to the AI SCIP page for further reading on the topic, so I'm quite confident that the $705m and $890m figures represent the same project. back

When I was initially trying to find information on Steelport after reading the National Observer article, and before being told about the DRAC data centre, I estimated the footprint of a 180 MW data centre based on existing facilities. The ballpark number I came up with: 300,000 square feet - around 7 acres. This feels like it might be notable. back

According to a recent piece by Ed Zitron, nobody has successfully completed a 1-gigawatt data centre yet. Surprising, I know! Several projects have completed some *portion* of a gigawatt data centre, and have therefore been reported as being operational, but Ed is unable to find an example of a single one that has actually been completed. back

Slate, I'm going to stress here, are not folks who I'd expect favours from just because I asked them nicely. An entire wing of their corporation is called the "Grocery REIT" division, [archive link] whose business model has been to buy up distressed grocery store locations for cheap - mostly early in the pandemic - and then "roll up our sleeves and find ways to increase rents". In other words, a purely extractive process where they make money by increasing the price of everyone's groceries while creating precisely zero social benefit. back

Which raises a question about the DRAC data centre: instead of the feds spending $705 million on research chatbots - with the bulk of that money going toward Nvidia GPUs - why not instead spend the cash on desperately-needed public housing, keep those dollars in Canada, *and* create jobs for tradespeople that last a lot longer than 18-24 months? back

I'm unable to find any mention of data centres in any of Slate's press releases or news articles covering Steelport prior to 2025. So it really does seem like a company that's getting desperate. Incidentally, the site has also been divided up into parcels which are for sale, if anyone's interested. back

The Wonder Valley Files, Part 1: Big Promises, Suspicious Numbers

2026-04-09T00:00:00Z

Thanks to a cache of documents obtained late last year through Freedom of Information requests made by Canada's National Observer, Canadians can now gain a little more insight into Kevin O'Leary's "Wonder Valley" project.

But the heavily redacted documents, which have been examined by Blunder Valley, might raise as many questions as they answer. We'll be exploring some of these questions in a series we're calling "The Wonder Valley Files".

Swimming with sharks

"Would be the largest data centre in the world", reads the punctuation-free email. You can almost hear the breathless, barely-contained excitement from its sender, Kyle Reiling, who's been hoping to land a big-fish project like Wonder Valley for years.

In December 2025, Reiling, director of the Greenview Industrial Gateway (GIG), told The Logic that the Municipal District of Greenview has spent "$70 million to date" on work to get the industrial park shovel-ready. But despite those efforts, and despite the excitement around the Wonder Valley announcement in November 2024, there hasn't been a whole lot of visible progress since: other than a single road with a truck turn-around, the site remains empty.

And after combing through this 600-page cache of documents and emails, I think it's worth asking if Shark Tank personality Kevin O'Leary was ever serious about building the "largest data centre in the world", or if he always had other plans.

O'Leary, a showman whose actual business ventures have been perpetually mired in controversy, built his career on being the center of attention, and the Wonder Valley proposal definitely got a lot of press. But O'Leary Ventures isn't a data centre company. Why would any smart investor hand them $70 billion to build a data centre campus several times the size of anything already in existence?

(So far, it seems that the only investor interested is US billionaire Frank McCourt, whose "Project Liberty" previously partnered with O'Leary on the astroturfed "People's Bid for TikTok", a doomed attempt to purchase the social media platform. McCourt has pledged to contribute $100 million, or 0.14% of the total Wonder Valley price tag.)

So is Kevin O'Leary actually about to build the world's largest data centre? Here's a more plausible scenario: O'Leary made a bombastic promise that he had no intention of keeping, and he's been misleading the municipality and the province in order to advance his own interests.

To be clear, that doesn't mean he won't build anything on the GIG at all. Perhaps the real plan was to construct a much smaller (but still enormous) AI data centre—or perhaps, as some of the emails hint at, a huge bitcoin mine instead.

Playing them like a fiddle

Regardless of O'Leary's true intentions, making a wildly oversized pitch has its advantages. For one thing, everyone in the industry is guaranteed to hear about it, which makes it easier to find companies eager to work with him.

And if there's a favourable political environment, this strategy can save a lot of headaches in terms of planning, permits, and so on: O'Leary surely understands that if local and regional governments are sufficiently excited about a project, they'll jump at the chance to clear away any regulatory hurdles. A smaller project is unlikely to get this kind of white-glove treatment.

O'Leary may also be leveraging the municipality's desperation. One email in particular, sent to the municipal planner in December 2024 by a concerned member of the public, doesn't mince words:

"I know business is business but he is taking you, this Province and The Premier as fools, don’t underestimate the timing of all of this as he thinks we are on our knees and suckers."

Greenview and the GIG have had multiple proposals for the site fall through over the years, even after sinking so much time and money into it. Reading through the emails, you get the sense that everyone on the municipal side is a bit star-struck by O'Leary, and actively wants to believe what he's saying, even to the point of ignoring multiple red flags.

For example, the municipality's top elected official, Reeve Ryan Ratzlaff, says he's not worried about delays or plans falling through, telling The Logic, "they [OLeary's team] just started moving a little bit slower to make sure that they had everything as perfect as they could get it”.

All of this has worked in O'Leary's favour: both Greenview and the GIG have put in real effort to accomodate his demands, including pushing through subtly-worded but important changes to zoning rules on behalf of the project (something we'll discuss in a future piece). Danielle Smith's Alberta government, for their part, recently exempted Wonder Valley from requiring an environmental assessment. And federally, Mark Carney's Liberals are now allowing gas-powered data centres in Alberta, despite the fact that this project would single-handedly set Canada's emissions goals back 20 years, reversing all the progress from phasing out coal in Alberta.

A Suspiciously Convenient Water Usage Number

One of those aforementioned red flags appeared in an email sent by Reiling on October 31, 2024:

"My water numbers were wrong

Just so you know

6GW is around 20 million cubes of water".

One day prior, Reiling emailed similar news to the company handling the water supply infrastructure for the GIG, telling them he'd "just received confirmation that their water requirements at full build out is [redacted] so this design is perfect."

Here's the problem: according to an expert estimate reported in The Energy Mix, the water consumption of a data centre of this size, plus the on-site gas plants that would power it, is on the order of 112 and 195 million cubic metres per year. That's 5 to 10 times larger than the number Reiling was given. That expert estimate is also roughly in line with Blunder Valley's own estimates of the likely total, based on available published data.

Now, maybe someone's finger slipped while they were punching numbers into their calculator, and that's where the 20 million figure comes from. Simple mistake.

But if it is a mistake, it's a remarkably convenient one.

Because it just so happens to be a little less than 24 million cubic metres—the maximum amount the GIG is allowed to divert from the Smoky River each year, according to the Conditional Master Water License issued by the province. 24 million is also the number that the water intake & storage designs were being based on.

20 million, in other words, would be the perfect number for O'Leary Ventures to feed to Kyle Reiling if, say, they needed to soothe any last-minute, pre-announcement jitters about the overall feasability of the project. Probably just a lucky accident though!

Malice, or incompetence?

So what's going on? Here's few possibilities:

Possibility #1: the "20 million cubes" number really is the result of a massive calculation error.

If true, it suggests a level of incompetence that shouldn't inspire confidence about the future of the $70 billion project. The physical reality is that data centres and gas plants require a huge amount of water for cooling; without it, things overheat and break.

Possibility #2: the "20 million cubic metres" number is a lie, and Wonder Valley plans to divert and consume far more water than their permits allow.

This seems risky. For one thing, it sounds wildly illegal. But the risk here goes beyond legal liability, or even moral hazard: it's a risk to other people's lives and livelihoods. As reported in The Energy Mix, in July 2025 the Municipal District of Greenview joined Pincher Creek in declaring an agricultural disaster for the livestock industry due to "worsening drought conditions and persistent moisture shortages".

And as we noted in a previous piece about a data centre proposal in Pincher Creek, Alberta rivers, particularly those fed by the receding glaciers and snow pack of the Rocky Mountains, are expected to have their flow rates decrease as the world gets warmer. Whether we like it or not, we need to start making difficult choices about what kinds of human activity we want to prioritize.

Will we give U.S. tech giants first dibs on increasingly scarce water so they can continue pumping out AI slop 24/7/365? Or should we instead prioritize things like, say, food production?

Possibility #3: 20 million cubic metres of water is only for the data centres, and doesn't include any gas-fired power generation.

Even then, the math doesn't really work. Although the data centres themselves would consume less water than any gas plants built to power them, we estimate a water consumption of around 65 million cubic metres per year for the data centres alone.

Regardless, let's assume this is how they came up with their number: the data centres still need to get power from somewhere.

But from where? The already-strained Alberta grid?*

Possibility #4: The number is accurate, but only because the Wonder Valley project was always meant to be much smaller than O'Leary claimed.

In this scenario, the question becomes, "what size of data centre would consume 20 million cubic metres of water per year?" And the answer, based on the available water consumption statistics for AI data centres and gas-fired power generation, is a 1-gigawatt facility, give or take.

Again, to be clear: data centres on this "smaller" scale are still absolutely massive, and only a handful have even been built. It's just not quite as headline-grabbing as what O'Leary pitched.

More math, more problems

Another unanswered question: how much is this project actually supposed to cost? Recent reporting suggests the cost of a 1-gigawatt AI data centre is around $25 to $35 billion. Yet O'Leary is publicly pitching a $70B price tag for a project several times that size? Either costs have tripled in the last 18 months, or the Wonder Valley numbers were pulled out of thin air.

And lastly, according to an April 2025 email to Kyle Reiling from Fisheries and Oceans Canada (obtained last week through additional FOI requests made by Canada's National Observer), the federal regulator has only approved the annual diversion of 6 million cubic metres of water per year from the Smoky River, using a temporary, seasonally-installed pump, until 2033. In other words, just a quarter of the volume that folks at the GIG are still claiming the project has permission to use.

Interestingly, this very same email appeared in the earlier cache of documents released by the municipality, but had been fully redacted.

So what is Kevin O'Leary's real plan? It's still unclear. One thing's for sure: no one should trust a single word that comes out of his mouth.

* Part 2 of The Wonder Valley Files will look at the question of power generation.

Is this greenwashing? Asking for our friend Kevin O'Leary

2025-07-03T00:00:00Z

Recently, we wrote about how Paul Palandjian, CEO of O'Leary Ventures, admitted in April that carbon capture is too expensive for hyperscalers, telling conference attendees, "all of these green aspirational stories are great, but it's not happening yet, and we're all trying to figure out how to lower that carbon footprint and do it cleanly".

And yet, an infographic [archive] uploaded in June to Wonder Valley's website claims that the project will be "the world's most sustainable data center campus" and implies a zero-emissions system with "full carbon capture".

Are they kidding?

Canada's anti-greenwashing rules

One might wonder if O'Leary Ventures' legal theam has anyone keeping tabs on Canadian legislation - specifically the amendments made in 2024 to the Competition Act that added langage targeting greenwashing.

If not, they might want to start, because the financial penalties for engaging in deceptive marketing practices can be pretty hefty by most people's standards: up to $10 million for a first offense ($15 million for subsequent violations), 3x the benefit derived by the company due to the deception, or 3% of the company's total annual revenues, whichever amount is greater.

The rules themselves are designed to be applied quite broadly - not just to the marketing of consumer products, but also to environment-related claims about business activities in general. And it's working. Oil & gas sector companies have suddenly become a lot more careful with their marketing material: in one case, as the rules were about to take effect, the Pathways Alliance decided to take down their website entirely [archive] rather than risk prosecution.

Incidentally, when Mark Cameron, Pathways Alliance spokesperson and former top advisor to Stephen Harper, appeared in front of the Senate Committee on Banking, Commerce, and the Economy in May 2024 to oppose the amendments, he was asked a number of questions about carbon capture and storage. One startling admission he made was that the cost of carbon capture is "probably in the $200 to $250-a-ton range". For comparison, Canada's future price on emissions in 2030 will only be $170 per ton!

The Pathways Alliance were right to be worried, because the Competition Bureau doesn't mess around. In 2022, two years before the greenwashing amendments were passed, Keurig Canada paid settlements totalling nearly $4 million for claiming their coffee pods are recyclable, despite only a few recycling programs outside of BC and Quebec accepting them. (Also, there are apparently a lot of steps that need to be taken in order for a coffee pod to actually get recycled.) In other words, Keurig's claim may have been technically true - or true in theory - but it was false in practice.

Palandjian's comment in April about "green aspirational stories" shows that, like the Competition Bureau, he recognizes there's frequently a gap between marketing narratives and objective reality. In Keurig's case, this gap cost them $4 million - not exactly chump change. By the way, this was the penalty with Keurig Canada voluntarily cooperating with regulators.

Is Kevin O'Leary hoping for a fight with the feds?

So, given the no-nonsense enforcement approach of the Competition Bureau, it's a little odd that O'Leary Ventures is so brazen about its environmental claims. It almost makes you wonder if, just maybe, they're hoping for a fight?

We've already seen that Premiers Danielle Smith and Scott Moe are extremely keen to have a brawl over federal environmental regulations: they recently made a joint statement containing a slew of demands, which include repealing the Impact Assessment Act, removing the ban on oil tankers on the northern west coast, and other minor tweaks like "repealing any federal law or regulation that purports to regulate industrial carbon emissions, plastics, or the commercial free speech of energy companies".

(For anyone not familiar: that last phrase - "commercial free speech" - is a euphemism for making false and/or misleading environmental claims. It means, in other words, lying to the public. Greenwashing.)

As for Kevin O'Leary, it wouldn't be surprising if he's hoping to use all this as a marketing opportunity. Inserting himself into the fight would be a great excuse for him to get back in his comfort zone, and do another US media tour - where he can once again spew BS about Canadian regulations with zero pushback, while simultaneously promoting his US business interests. For someone like O'Leary, who's frantically pitching projects with price tags in the billions, $10 million might sound like the cost of doing business.

Fighting greenwashing

Ultimately though, the issue isn't whether or not O'Leary is aware of Canadian regulations, or whether or not he wants to create a media spectacle for his own benefit. The real issue is that there are unambiguous environmental claims being made on the Wonder Valley website that are directly contradicted by the words of his company's CEO.

Maybe someone oughta ask the Competition Bureau if they plan to do something about that?

Brookfield Provides $750M Credit Facility to Peter Thiel-Funded AI Data Centre Company Crusoe Energy

2025-06-30T00:00:00Z

Palantir co-founder's VC fund led a $600M funding round for Crusoe just months ago. Crusoe wants to build new gas-powered data centres in Alberta.

'Wonder Valley' is far from the only proposal for a massive, gas-powered AI data centre in Alberta. US-based startup Crusoe Energy also hopes to get in on the fossil fuel burning free-for-all, and they're getting help from some big names.

In December 2024, for example, Crusoe raised $600 million in a Series D funding round led by none other than Founders Fund, Peter Thiel's venture capital firm.

Thiel, as many readers will already know, is co-founder of Palantir Technologies, the data analytics company providing software to help ICE kidnap civilians in the US for Trump's mass deportation program. Despite this, the President of Palantir Canada, David MacNaughton, is a member of Prime Minister Mark Carney's Council on Canada-U.S. Relations.

Then in March, Crusoe was handed $225 million in debt financing in a round led by NY-based firm Upper90. This round also included funding from the British Columbia Investment Management Corporation (BCI), which manages investments on behalf of BC's public sector pension funds. Peter Milburn, Chair of BCI's Board of Directors, was the province's Deputy Minister of Finance until being appointed to this position at the end of 2016 by BC's Minister of Finance.

This all might sound like an astonishing amount of money - an amount that could keep virtually a startup going for years. But Crusoe Energy, like so many other AI-related companies, is burning through its stacks of cash at an even more astonishing pace. And on 11 June 2025, Brookfield Asset Management provided Crusoe with access to a $750 million credit facility.

What is Crusoe doing with all this funding?

Crusoe Energy, which started out as a Bitcoin mining company, is currently building the second phase of a 1.2 gigawatt AI data centre facility in Abilene, Texas. The site, called the Lancium Clean Campus, features "backup" power generated by five on-site gas turbines, although it's not clear what proportion of its energy usage will ultimately come from what sources. The 'Clean Campus' branding remains, however.

As Bloomberg has reported, the ability to quickly construct facilities like this, connect them to the grid, install gas turbines, you name it, is in large part due to a lack of regulation of energy projects in Texas. And while this deregulated environment is attractive to companies like Crusoe, with their immense piles of cash and a desire to move fast and break things, the lack of public oversight comes at a heavy price: in 2021, six people died in Abilene County as a result of widespread blackouts during a severe winter storm. At least 51 others died in other parts of the state, and the property damage incurred due to the power outage is estimated to have cost Texans $195 billion.

Crusoe's Alberta plans

Crusoe Energy markets itself as having a "deep commitment to scaling AI infrastructure in a climate-aligned way", and uses "Clean Campus" branding to mask the fact that their Texas operations are powered, at least in part, by fossil fuels.

In keeping with this deep commitment to greenwash their image, Crusoe signed a 'framework agreement' (read: not a final contract) in February 2025 with Kalina Distributed Power, an Australian-owned gas plant company promising to build three 170 megawatt gas plants providing "affordable natural gas-fired power configured with CO2 capture and sequestration." But as we recently heard from O'Leary Ventures' CEO Paul Palandjian, carbon capture systems, even if they can reliably sequester carbon, are still non-viable in situations like these, because they increase the cost of the generated electricity to the point where data centre clients are no longer willing to pay for it.

And at least for the time being, even Danielle Smith is keenly aware of the dangers of an over-extended grid, and is reluctant to allow the construction of hyperscale data centres unless they supply their own electricity generation.

Three paths forward

So there are a few possibilities.

One is that Crusoe and Kalina will cancel their plans for Alberta data centres, despite having already acquired three parcels of land totalling 545 acres, and having conducted "environmental desktop and biophysical studies" at two of the three sites so far.

A second scenario is that the federal and/or provincial government could decide to provide AI-related projects with additional rebates and/or tax breaks if they implement CCS, beyond what they already provide (at present, the federal government provides a 50% rebate on the capital cost of carbon capture infrastructure, while Alberta kicks in 12%). Of course, this might be seen by Canadians as a bailout for Thiel and Brookfield using public money.

A third possibility: Crusoe and Kalina will forge ahead, build the gas plants and data centres, but abandon the plans for carbon capture & storage. This would line up with what's being proposed for another Kalina project in Alberta, where CCS would "possibly" be implemented “as legislative and commercial circumstances warrant.”

Who knows what will happen, but one thing's for sure: we should definitely maintain a healthy skepticism when evaluating the promises being made by companies looking to build these massive data centres.

Wonder Valley team admits it: Carbon Capture is Too Expensive for Hyperscalers

2025-06-03T00:00:00Z

Kevin O'Leary's Wonder Valley business partner Paul Palandjian publicly concedes that CCS was never really going to happen, promises to 'crush' competition

Kevin O'Leary spent the early part of 2025 visiting Mar-A-Lago with Danielle Smith and telling U.S. media a completely fabricated story that half of Canada is "very interested" in becoming the 51st State. For a brief moment, he was quite possibly the most hated person in Canada.

This brutal PR misstep might be why O'Leary is currently keeping a low profile (at least by his standards) while the planning for his Wonder Valley project quietly chugs along.

In the meantime, Paul Palandjian has been doing the conference panel circuit, including an April appearance at an event called 'West Palm Beach: City of Ideas', organized by Michael Greenwald, head of Financial Innovation and Digital Assets at AWS.

Palandjian, O'Leary's right hand man at O'Leary Ventures, took part in a discussion moderated by Helima Croft, Head of Commodity Strategy at RBC Capital Markets.

The panel, called "Private Investing and the Future of Data Centers and Energy" also included Kathleen Barrón, a VP at Constellation Energy Corporation (the same company that's working with Microsoft to reopen Three Mile Island!), and Adam Rodman, founder & CIO of Segra Capital, who bears an uncanny resemblance to Patrick Bateman, and whose firm is heavily invested in uranium mining and nuclear power.

While the full video isn't (yet?) available, an April post on the Wonder Valley website links to a clip on O'Leary's YouTube channel. I've uploaded a copy to a new YouTube channel just in case.

One small error that Palandjian makes: he claims that the Wonder Valley site is sitting atop "200 trillion metric tons of natural gas" rather than the previously quoted figure of 200 trillion cubic feet. If any of the other panelists noticed the mistake, they certainly didn't let on.

The Big Blunder

But the big blunder that Palandjian made was tactical: he told the truth about the non-viability of Carbon Capture and Storage.

Prior to this, the Wonder Valley team (and the Greenview Industrial Gateway team, and the Alberta Government) had been floating CCS as a way for this and other similar projects to reduce or even eliminate their carbon emissions. Turns out they know it's too expensive to be viable:

"[Carbon capture] is not ready for prime time yet; the efficiency loss when you capture carbon off these gas turbines is massive. So if you're generating 1.4 gigawatts of natural gas power, you'll lose 40% of it just sequestering carbon. And what Michael [Greenwald of AWS] and Amazon cares about is net IT load, you've gotta be able to give them power. So they're out in the market buying carbon credits; the running joke in the data centre business is, the only thing they're recycling is the cups inside the data centres, so, you know, all of these green aspirational stories are great, but it's not happening yet, and we're all trying to figure out how to lower that carbon footprint and do it cleanly. And most people agree that nuclear is the way out, hopefully sooner rather than later."

In other words, the Wonder Valley project will not be sequestering carbon, because it makes the electricity cost more than hyperscalers like AWS are willing to pay.

"We're gonna crush the competition"

Palandjian continued with a rather ominous message:

"I will boldly say the we are gonna win this war, we're gonna crush our competition - but we have the most talented people, we have the most sophisticated financial system and capital markets; I don't see a world in which we could lose, but it's time to act, now."

From this clip alone, it's unclear who the competition is that Palandjian is referring to. However if I were him, I'd be most worried about opposition to the project from local residents, First Nations, climate activists, and others concerned that Wonder Valley could pump obscene amounts of carbon into the atmosphere, and put pressure on water-availability in an already drought-stressed part of the planet that's dealing with massive wildfires on a regular basis.

AI makes us less human

2025-03-08T00:00:00Z

AI systems take away opportunities for us to do what we do best: listen, empathize, and make caring, thoughtful decisions.

The narrative

The next time you hear a tech CEO or a politician saying that AI will allow us to do all sorts of amazing things, that it represents a 'fourth industrial revolution', that it allows us to streamline and optimize government operations, or that it's going to transform the delivery of medical care - and therefore we need to devote huge amounts of public money and resources to it - ask yourself this: where's the evidence to support this?

Major tech companies have already poured several hundred billion dollars into buying chips, building data centres, burning electricity and draining freshwater resources. Has your life improved in any meaningful or transformative way?

What is AI good at?

What is AI good at? Well, computers tend to be a good choice to perform tasks that can need a lot of calculation, but that don't require any understanding of the world. AI is no different. Some AI systems are built for those types of applications; AlphaFold, for example, models the folded shapes of proteins, produces quite accurate results, and may prove to be very useful for medical research.

DeepMind, the group that developed AlphaFold, has also built systems that try to tackle very different kinds of problems. One of these is AlphaGo, which plays Go, a 2500-year old game. The rules of the game are simple, but the game is difficult to master - and Go's large 19x19 board (compare that to an 8x8 chess board) means there are a lot more possible moves that a computer has to consider.

And sure enough, it wasn't until 2016 that the most advanced Go-playing computers began consistently winning against the world's top-ranked human players. By that point, Chess computers had already been beating world champion grandmasters for a couple decades. Getting to this point with Go was seen as a major accomplishment.

Since then, presumably, Go-playing computer systems have only become more dominant ... right?

“As a human it would be quite easy to spot" - Kellin Pelrine

As it turns out, even these computational juggernauts have glaring weaknesses.

Go is played by taking turns placing stones on the board's 19x19 grid, and attempting to surround groups of the opposing player's stones, in order to remove them and capture territory.

So you'd think that an AI system like KataGo (widely considered to be on par with the AlphaGo system that beat top-ranked Lee Sedol in 2016) would be able to spot its opponent doing something really obvious like, say, placing their stones on the board in a big loop around the computer's pieces.

But this pretty much describes how Kellin Pelrine, a highly-ranked amateur Go player and professional AI researcher, beat KataGo in 2023.

The victory wasn't a one-time fluke, either: Pelrine won 14 out of 15 games against KataGo using the same strategy. Which makes it very clear that no matter how clever these AI systems might sometimes appear to be, they don't actually 'know' or 'understand' anything. When it comes down to it, they're still just calculators.

To expand on this: an AI system might be capable of doing a specific task fairly reliably, but what it can't do is generalize its problem-solving approach and apply it to different problems. In fact, you could argue it's even worse than that: as Pelrine's victory showed, these AI systems don't have what we'd call 'common sense' or 'insight' even when you're only looking at the extremely narrow task the AI is specifically built to perform.

Unthinking, unfeeling

This lack of common sense can have consequences that are a lot worse than a string of losses in a board game.

In the Netherlands, for example, an algorithmic system incorrectly flagged thousands of Dutch families as having fraudulently received child welfare benefits. Even something like forgetting a signature on a document, or having dual citizenship, could result in a family being flagged as fraudsters, and being forced to pay back tens of thousands of euros they had (legitimately!) received over the course of years.

These mistakes destroyed lives. Families were torn apart, people lost their homes, many spiraled into depression. These people were the casualties of a push for 'government efficiency' at the expense of empathy, care, and basic common sense & decency.

Automation bias

Ok sure, you might say, algorithms don't have common sense or empathy - but what if there'd been a human being double-checking the system's outputs to make sure it wasn't making tragic mistakes?

Well, we happen to have scientific studies and real-life examples of AI deployments where there were humans in the loop, monitoring and double-checking the systems' outputs, but unfortunately, it doesn't seem to have helped a whole lot.

In one scientific study, titled Automation Bias in Mammography, researchers rigged an AI to occasionally mis-categorize a mammogram. The AI would show its assessments to professional radiologists, after which the researchers asked the radiologists to categorize the mammogram. It turned out that it didn't matter whether a radiologist was fresh out of college or had decades of experience: the accuracy of their diagnoses dropped dramatically in the cases where the AI was wrong.

The explanation? Human beings have a tendency to over-value decisions made by automated systems, to the point where we'll stop trusting our own skills and knowledge, and defer to the machine.

Luckily, those misdiagnoses happened in a controlled setting, and no patients' lives were affected. But there are plenty of examples of automation bias leading to significant real-world harm, too.

For instance, there are multiple cases where "100% match" results from facial recognition technology (archive link) have led U.S. police to wrongfully arrest people.

In many of these cases, there were basic facts that should have made it clear that people were completely innocent, even before they'd been arrested. Things like: having different tattoos, being obviously pregnant despite no eyewitness statements having mentioned this, or having been at work in a different state when the crime was committed. In some cases, it took over 2 years for people's names to be cleared.

At this point, it seems like it might be worth asking: when it comes to using AI in medical diagnosis, or using facial recognition in policing, are we sure these things are really a benefit to society? Or are they actually making us all less healthy, and less safe?

Offloading responsibility

If it wasn't bad enough that automation bias makes people place too much faith in these systems, imagine the same thing but one step further: deliberately implementing AI for the purpose of offloading moral hazard and responsibility.

Suppose, for example, you're running a health insurance company. You sell policies, but it's in your financial interest to pay out as few claims as possible without losing too many customers. It's also in your financial interest to keep your payroll as small as you can.

What's the solution? Automate things! That way, when you deny a claim and someone complains, the overworked customer service rep has no reason to look into it further, because the complex computer system has already made an unbiased determination based on all the available information.

Of course, as we've already seen in the previous examples, these systems aren't actually good at making reasonable judgements - but that hasn't stopped insurance companies from introducing AI into their claims review processes.

In the US, there are currently 11 states that have either introduced or passed legislation that limits, or even outright bans, the use of AI for reviewing health insurance claims.

Obscuring discrimination

Automated systems can also offload responsibility (and make legal liability harder to determine) by hiding discriminatory practices and bias behind algorithmic complexity.

For example, a bank might be barred from collecting data on their customers' race or ethnicity when determining who's eligible for a business loan, but might still find a way to use things like postal codes, perhaps in combination with other data points, as a proxy for race and/or ethnicity.

This makes it incredibly hard to detect discriminatory practices. Particularly when it comes to AI systems, it's nearly impossible to tell from the outside how a computer made a decision.

In other words, not only can the bank discriminate against people, they can also do so while claiming that their systems are unbiased - after all, they don't collect data on race and ethnicity! The computer takes a whole bunch of input data, puts it all in an opaque box, and spits out a binary yes/no determination that can't be argued or reasoned with ... or really even understood by the person working at the bank who's passing on the good or bad news.

Holding AI systems accountable

I feel like there's really only one conclusion: we should never allow AI systems to make judgement calls that directly affect people's lives. Someone wants to simulate protein folding? Go for it. Explore board game strategy? Be my guest. But AI should never be allowed to make decisions about things like people's healthcare, their access to government programs or services, or the likelihood that they're guilty of a crime.

So next time you hear a politician or business leader talking about all the amazing things we're going to be able to do with AI, ask them: what are these amazing things, exactly?

"Never Going to Get Built": are the plans for Wonder Valley evaporating?

2025-02-23T00:00:00Z

Opposition is growing, and the case for massive AI data centres is getting weaker - was 'Wonder Valley' always just a mirage?

“I’m not at all concerned that this project will end up being a horrible thing because this project is never going to get built."

- Alberta NDP leader Naheed Nenshi, Feb. 14, 2025

Even back when it was announced in December, some observers couldn't help but wonder if Kevin O'Leary's 'Wonder Valley' was actually a serious project. The project would be nearly an order of magnitude larger than anything that's come before it, the promotional material seemed like it was slapped together in a rush, and the novelty of generative AI was already starting to wear thin.

Even still, it's important to take this sort of proposal at least somewhat seriously, and push back. Thankfully, that's been happening on several fronts.

Sturgeon Lake correctly demands substantive discussion

The most important and effective opposition has come from Sturgeon Lake Cree Nation, whose Chief Sheldon Sunshine rightly questioned the complete lack of consultation from both Kevin O'Leary's team as well as Alberta's UCP government in a January letter to the Premier.

And quite frankly, it seems fair to describe the Province's response to Chief Sunshine as insulting. Even a one-hour in-person meeting, which was understood as what was being offered, feels like a token gesture, considering it's ostensibly meant to be a nation-to-nation discussion about a $70 billion mega-project. But to downgrade this to a half-hour Zoom call? I applaud Sturgeon Lake for cancelling the meeting.

"On its face, it seems that this meeting is being set up to provide a tick on the box so that the province can say they've fulfilled their duty to consult", wrote Chief Sunshine in a second letter on Feb 12, as reported by CBC Edmonton.

Of course, now the Alberta government is "working to coordinate an in-person meeting", according to CityNews Edmonton, in an update posted February 18.

Naheed Nenshi knows what's up

The Alberta NDP leader wasn't pulling any punches in a press conference on Feb 14th. In addition to declaring the project a non-starter, he noted that Premier Danielle Smith clearly over-promised what she could deliver to O'Leary, falsely claiming she could issue permits, in exchange "for a ticket to Mar-A-Lago".

Nenshi also pointed to the impact of DeepSeek, an AI model that can do the same things as the leading US models, but consumes a fraction of the energy. It'll be difficult for tech companies to justify spending hundreds of billions more on AI infrastructure if they still can't figure out how to turn a profit at a tenth of the cost.

O'Leary may have already pivoted to other things

It's confusing: Kevin O'Leary seems to have been his own worst enemy when it comes to rallying support for the Wonder Valley project. His TV appearances, where he claimed that Canadians are "very interested" in hearing about a proposal to form an "economic union" with the US - where we would give up control of our own currency, and with it our sovereignty - haven't gone over well in Canada. Neither did his trip to Mar-A-Lago, nor any of his other attempts to ingratiate himself to President Trump while throwing Canadian interests under the bus.

And while O'Leary might not be the sharpest tool in the shed, he might have a rationale for burning his bridges with Canadians. Ultimately, O'Leary is a spokesperson and and opportunist, not someone with a solid business plan. And the original Wonder Valley proposal, with its wildly ambitious scale, well, maybe it was always more of a negotiating tactic, a way to get some leverage elsewhere?

For example: maybe it was something he could show local and state governments in North Dakota and West Virginia - places where O'Leary has said he would like to build data centres - to make them scared they'll lose out on some kind of amazing, can't-miss opportunity to build enormous computational warehouses that will drain their rivers and energy grids.

And later on, when he was going on Fox News as a self-appointed emissary to Canada on behalf of Trump and his annexation ambitions, maybe O'Leary's primary goal was getting into Trump's good graces in the midst of the TikTok kerfuffle. O'Leary, you see, had been hired to promote something called "The People's Bid for TikTok", something that wasn't really a grassroots effort, but really a play by US billionaire Frank McCourt to take control of the social media app's US operations.

McCourt's 'Project Liberty', which is behind The People's Bid, runs a blockchain-based social media app called 'MeWe', which never really got off the ground. And unfortunately for both McCourt and O'Leary, neither did their TikTok offer.

Poor Kevin, he just can't seem to catch a break!

Danielle Smith and the UCP are drowning in scandals

Lastly, if the Alberta government had been thinking they could skirt treaty obligations and/or regulations in order to get projects like this built, you have to figure they aren't thinking that now, given the intense scrutiny over healthcare procurements that are looking worse every day. Tough for any other sketchy deals to fly under the radar once the public's trust has been so thoroughly broken.

Ultimately, it's impossible to know for sure if Kevin O'Leary was ever serious about Wonder Valley. And maybe he never really was. But as both O'Leary and the UCP have shown so clearly, you can't give these folks an inch.

Bread Price Fixing? Pfft. The Real Innovation is AI-enabled 'Tacit Collusion'

2025-02-08T00:00:00Z

image by lyzadanger. Used under a CC-BY-SA 2.0 license.

How big grocery chains use AI and data analytics to boost profits and drive out competition

When we're talking about massive AI data centers, it's not just the enormous amount of energy, water, and mineral resources that we should be concerned about. We also need to ask, what is the actual purpose of all this computation? Who does it benefit?

Who benefits from all this computation?

Broadly speaking, an AI system is something that uses a huge amount of data and/or instructions in order to produce guesses and predictions. Maybe it guesses what you mean when you ask for an image of Pope Francis wearing a puffy jacket, or maybe it tries to predict the next word you type into your phone.

Now, it's not exactly true in every case, but for the most part, the more data you feed into the system, the better its guesses and predictions will be. In the case of the Pope image, the data set is made up of nearly every image the AI company could find on the internet, whether or not they had permission to use it. In the case of the word prediction, it might be a combination of some basic rules of grammar, words you've used before, and a huge amount of example text.

But it's hard to justify spending hundreds of billions of dollars on hardware and data centers simply in order for people to generate goofy images, or to save you a few keystrokes when you're texting a friend. Where's the real business case for this stuff? Who's going to pay the big bucks to make use of all that infrastructure?

Well, if you have a bunch of data that no one else has, and you feel like you can gain a competitive advantage by using it, then you're an ideal AI customer!

Rewards programs = data collection

Major grocery chains have been collecting a ton of data on their customers' spending habits for years through loyalty cards. These are huge data sets - and quite often, no one else has access to them. So then, what if a grocer could use all that data to their advantage, by generating guesses and predictions about how customers are likely to respond to promotions or price increases? What if a supermarket chain could monitor their competitors' prices, and automatically change the prices on their own shelves in response?

What if (and to be clear: this is not speculation, retailers in the US are actually talking about doing these things!) facial recognition could allow a company to gather even more data on you, guessing your age, gender, even your current mood, linking this to your purchasing history, and then changing the price tags on shelves in real-time as you walk up to them? Imagine, for example, an algorithm that hikes the price you get shown for tampons, on the days when data from your period-tracking app suggests you might urgently need them!

AI hurts competition

Here's where we start to see exactly who is most likely to benefit from all this computation: the companies that are collecting the most data, and that can afford to spend the most on hardware & computation. In other words, these systems create advantages for the biggest companies, and help them drive out competitors and consolidate market power.

In many ways it's a self-reinforcing cycle. People struggling to cope with increased food prices will naturally gravitate toward loyalty card programs offering discounts and rewards. Who can blame them! But every time someone uses one of those cards, the company collects more data to feed into its system. Kroger, one of the biggest US grocery retailers, says that 96% of its sales are from people who use the store's loyalty card. Is it a surprise to find out that Kroger owns a data analytics company, and received a letter in August from two US Senators investigating their use of data-driven 'dynamic pricing' and in-store facial recognition?

Power consolidation

Canada isn't very far behind on this stuff, either. Loblaws, Sobeys, and Metro are all in the process of switching to the use of electronic shelf labels that can be updated wirelessly, and honestly, they look so much like paper price tags that you might not have even noticed the change (I hadn't until someone pointed it out to me).

Even at the time it happened, there were people suggesting that the primary motivation for Loblaw's 2013 acquisition of pharmacy giant Shoppers Drug Mart was a desire to get their hands on the trove of customer data gathered by the Optimum loyalty card program, which already had 10 million members.

Cut to 2021: the program, now called PC Optimum, has data on 18 million Canadians, and Loblaw is hiring dozens of experts in AI and machine learning, while expanding its empire into banking, advertising, and even apps like PC Health.

'Tacit collusion'

Thankfully, Canada's Competition Bureau has started to look into the potential impact of AI systems on the prices Canadians pay for things like groceries and rent - things that we can't do without.

One possibility mentioned in their recent report is that AI algorithms could, in effect, coordinate to gouge consumers without leaving any direct evidence:

"There is a risk of AI or algorithmic pricing systems facilitating tacit collusion in which systems autonomously align prices without explicit human instruction, communication, or agreement.

Concerns were raised regarding accountability of the companies that implement these systems, and the ability of law enforcement to prove collusion in these cases." [link]

This hardly seems like a stretch: AI systems marketed toward retailers are designed to automatically adjust prices in response to what competitors are doing. We know this because they say so in the marketing material!

Now, the people promoting these systems like to claim that AI leads to lower prices through increased competitiveness. But if an algorithm is designed to maximize profit, then the real incentive is to charge as much as possible without losing customers. So, if the AI guesses that charging a few cents more than a competitor won't lead to customers driving an extra few minutes to the other place - well, it doesn't take long to see how this can result in prices creeping upward at both stores.

As the Competition Bureau report points out, this tacit collusion would all be happening within opaque algorithmic systems, and corporations could try to claim they had no influence over the process. It wasn't so long ago that Canada's major grocers were caught colluding to fix the price of bread. These same companies are going all-in on electronic shelving labels and customer data collection. Are we going to keep letting them get away with this stuff?

News recap, January 2025

2025-01-31T00:00:00Z

Photograph of a winter scene by Jorg Braukmann, and released under a CC-BY-SA-4.0 license.

Sturgeon Lake Cree Nation raises concerns about Wonder Valley, DeepSeek brings questions about the need for massive data centres, and Canada's Competition Bureau looks into AI's potential harms.

Sturgeon Lake Cree Nation writes open letter to Alberta's Premier Smith

The letter, sent on January 13th by Chief Sheldon Sunshine, points out that the proposed Wonder Valley project is on Treaty 8 territory and highlights the SLCN's reliance on the Smoky River, the body of water that the Greenview Industrial Gateway (GIG) is looking to draw from in order to cool the hyperscale data centres, as well as the gas plants that would power them.

(A previous Blunder Valley post attempts an estimate of how much water the project might consume, if fully built. The ballpark estimate is roughly ten times higher than the volume of water the GIG has requested a license to draw.)

The letter asks numerous questions about the proposal, and hones in on the "provincial government’s apparent coordination, behind closed doors and to the exclusion of our Nation, for a massive development on our traditional territory", and requests that the Province cease and desist from actions like these without their consent.

While O'Leary Ventures responded by offering to have members of their team meet with the SLCN, the lack of meaningful response from the Province indicates that the Wonder Valley proposal is on a much slower timeline, and has much less interest, than O'Leary has been suggesting.

For example, the CBC reports that the Alberta Utilities Commission, which must approve any large-scale electricity generation projects, "has not received an application related to the proposed development" yet.

Another tell-tale sign of the lack of investor interest in the $70-billion project: so far the only funding announcement has come from billionaire-owned McCourt Global, a firm that recently partnered with O'Leary in a bid to buy the US operations of Tik Tok. They have pledged CAD$100 million to the Wonder Valley project.

Release of DeepSeek AI model creates doubts about the viability of massive data center projects

January 20th saw the release of DeepSeek R1, a Chinese-made AI model that was trained at a fraction of the cost of models like o1 from OpenAI, but is said to perform at roughly the same level, is less expensive to run, and was released as open source software.

There is some irony here: it's likely because of export controls on the newest and most advanced chips that Chinese AI makers have put a huge amount of effort into increasing the efficiency of these systems.

Wall Street immediately punished American big tech firms for their feverish spending on hardware and energy, based on the narrative that advancements could only be achieved via scale.

Now that DeepSeek has put the lie to that notion, it's going to be a lot harder for companies to justify investing tens of billions more dollars into systems that have been struggling to find any path to profitability.

Does this mean that Wonder Valley is dead in the water? Time will tell - but Kevin O'Leary is clearly upset about DeepSeek.

Canada's Competition Bureau releases a report following their Consultation on Artificial Intelligence and Competition.

In 2024, the Competition Bureau sought comment regarding the potential impacts of AI systems on Canada's markets.

The resulting report, released on January 27 2025, mostly discusses competition among companies in the AI industry itself, but also mentions a number of areas where AI could cause significant disruption, or might lead to monopoly power.

Some of the concerns:

Algorithmic pricing can result in deliberate price-fixing and/or tacit collusion on prices. We've seen the former with RealPage artificially driving rent prices up, for example.
AI amplifying deceptive marketing practices and scams. From deepfakes to email targeting, the report lists multiple concerns about the impact of AI-enhanced fraud.
The advantages that AI gives large corporations. The expense of using AI systems to get insights and make predictions about customer behaviour tilts the playing field toward bigger companies. But for companies like Loblaw, who have also been collecting data on customers for years already, AI represents an enormous advantage. There's a real risk that these systems result in an even less competitive marketplace for Canadians, which is a pretty worrying thought.

Blunder Valley

Not Dead Yet: The Steelport Data Centre Fight Continues

A crucial first step

Slate will appeal this decision

Next Steps

Some More Juicy Tidbits

Lobbying

The PMO

Finance Canada

ISED

HICC

District Heating and other "Green" initiatives

Notes

The Steelport AI data centre(s): what we know so far

UPDATE, May 30:

UPDATE 2, May 27:

UPDATE, May 27:

Correction:

Background

Into the weeds

Part 1: The DRAC data centre

The numbers don't add up

The DRAC data centre building

Data centres are harming local grids

Other externalities

Water & heat

Noise

Data protection

But wait, this data centre is for research - That's good, right?

Part 2: Other Data Centres at Steelport

Local and regional concerns

Suggested actions

Notes

The Wonder Valley Files, Part 1: Big Promises, Suspicious Numbers

Swimming with sharks

Playing them like a fiddle

A Suspiciously Convenient Water Usage Number

Malice, or incompetence?

Possibility #1: the "20 million cubes" number really is the result of a massive calculation error.

Possibility #2: the "20 million cubic metres" number is a lie, and Wonder Valley plans to divert and consume far more water than their permits allow.

Possibility #3: 20 million cubic metres of water is only for the data centres, and doesn't include any gas-fired power generation.

Possibility #4: The number is accurate, but only because the Wonder Valley project was always meant to be much smaller than O'Leary claimed.

More math, more problems

Is this greenwashing? Asking for our friend Kevin O'Leary

The Wonder Valley website is making emissions-related claims that O'Leary Ventures' CEO has openly admitted are false. Are they running afoul of Canada's greenwashing legislation?

Canada's anti-greenwashing rules

Is Kevin O'Leary hoping for a fight with the feds?

Fighting greenwashing

Brookfield Provides $750M Credit Facility to Peter Thiel-Funded AI Data Centre Company Crusoe Energy

Palantir co-founder's VC fund led a $600M funding round for Crusoe just months ago. Crusoe wants to build new gas-powered data centres in Alberta.

What is Crusoe doing with all this funding?

Crusoe's Alberta plans

Three paths forward

Wonder Valley team admits it: Carbon Capture is Too Expensive for Hyperscalers

Kevin O'Leary's Wonder Valley business partner Paul Palandjian publicly concedes that CCS was never really going to happen, promises to 'crush' competition

The Big Blunder

"We're gonna crush the competition"

AI makes us less human

AI systems take away opportunities for us to do what we do best: listen, empathize, and make caring, thoughtful decisions.

The narrative

What is AI good at?

“As a human it would be quite easy to spot" - Kellin Pelrine

Unthinking, unfeeling

Automation bias

Offloading responsibility

Obscuring discrimination

Holding AI systems accountable

"Never Going to Get Built": are the plans for Wonder Valley evaporating?

Opposition is growing, and the case for massive AI data centres is getting weaker - was 'Wonder Valley' always just a mirage?

Sturgeon Lake correctly demands substantive discussion

Naheed Nenshi knows what's up

O'Leary may have already pivoted to other things

Danielle Smith and the UCP are drowning in scandals

Bread Price Fixing? Pfft. The Real Innovation is AI-enabled 'Tacit Collusion'

How big grocery chains use AI and data analytics to boost profits and drive out competition

Who benefits from all this computation?

Rewards programs = data collection

AI hurts competition

Power consolidation

'Tacit collusion'