top of page
Search

Voice of AI: 600 citizens stop a data center. For now.

My father would have asked if the factory would create jobs. My children ask how much water it uses. In Vancouver, some 600 people commented on a proposed AI data center, prompting the city council to postpone the hearing until after the October municipal elections. Meanwhile, AMD is acquiring a Toronto startup, Alphabet is restructuring its AI leadership, Meta is opening up a model that runs on a single graphics card, and a Vancouver-based company is promising to reduce its cooling water usage by 95 percent. What happens when computing power grows faster than public support?



My AI summer in Vancouver. And you're along for the ride. 🍁

Starting next week, I'll be reporting live from Vancouver for the next few weeks. I'll be meeting with founders, talking to providers on the AI frontier, catching up with old friends and former classmates at the University of British Columbia , and immersing myself in a scene experiencing both: an influx of capital that many European locations envy, and a level of public resistance that many European locations still face.

I'm not interested in the next evaluation round. I'm interested in the mechanisms behind it. Look forward to insights from an ecosystem that is three years ahead of us in many ways, and three years behind in others.


The scarcest resource is not compute. It is consent.

Vancouver's city council didn't reject anything. It postponed. That's the political term for a society that doesn't yet know what it wants. And it's happening in a province that's simultaneously investing heavily in computing power: The planned Telus data center downtown is slated to house more than 50,000 GPUs and provide up to 100 megawatts of power. BetaKit

The resistance is not a local phenomenon. In Montreal, an automated bakery was vandalized for the second time in just a few months, this time with the slogan "action anti tech." The hard data behind it is more unsettling than any spray paint: More than two-thirds of Canadians would oppose a data center in their neighborhood, and only one in three trusts AI at all . CNBC

Anyone who believes that acceptance arises automatically with the capabilities of the technology hasn't been reading the news for the last twelve months. That's precisely why every serious AI project doesn't begin with model selection, but with a sober assessment of the current situation. Our AI Readiness Check at readiness.oakai.de provides you with this assessment in under ten minutes, including whether your organization embraces the change or merely tolerates it.


The bill that nobody wanted to read aloud

In Vancouver, a company has turned this very gap into its business model. Wafr Technologies has raised 100 million of its targeted 300 million Canadian dollars to commercialize cooling technology for AI data centers. A typical data center consumes up to 10 million liters of water per megawatt per year, with cooling alone accounting for 30 to 45 percent of the electricity load. Wafr promises to reduce water consumption by up to 95 percent and cooling electricity consumption by up to 80 percent. The company came from the first cohort of the UBC Sauder Scale Up Program .

University, capital, and infrastructure problems are intertwined here. This is the pattern that makes the region interesting , and it is transferable.


Canada builds, America buys

The countermovement was evident that same week. AMD acquired Taalas, a Toronto-based startup that directly integrates AI models into silicon, aiming to eliminate the need for universal GPUs. Founded in 2023, the company had raised $169 million by early 2026. This means that four of Toronto's most promising chip startups have now been sold to or relocated to US corporations : Taalas , Untether AI, CentML, and Tenstorrent.

For European decision-makers, this is not an exotic, distant problem, but a reflection of their own situation. Those who fund talent and research but fail to ensure commercialization ultimately subsidize the value chain of others.

Vancouver shows that it can also work the other way around. IntelliStake, a Vancouver-based AI infrastructure company, is acquiring the Texas-based sensor technology company NanoAi for 17 million Canadian dollars in stock, tied to revenue and contract milestones. It's not the US corporation buying the Canadian startup, but the other way around. Meanwhile, Beedie Capital has secured a 50 percent stake in Vistara Growth and will become the anchor investor in a 500-million-dollar fund. The partnership dates back to two MBA classmates at UBC in the early 1990s. Networks that last for thirty years are no accident. They are a matter of strategic location. betakit



Two news items that will affect every AI budget

At Alphabet, Demis Hassabis is stepping down as operational leader of Google DeepMind, becoming chairman of the unit and also serving as Alphabet's chief scientist. On the same day, chief scientist Jeff Dean is leaving the company after 27 years to found Discovery Loop, a public benefit corporation in which Google has a stake. Sanjay Ghemawat, Oriol Vinyals, and Quoc Le are also leaving, and Alphabet's stock fell by around five percent. Radical Ventures from Toronto is the co-lead investor. Anyone who thought the concentration of AI power was complete is seeing the opposite. It is fragmenting along the lines of individual power. Search Engine Roundtable

The second piece of news is even more operationally relevant. Meta has released Muse Glimmer, a model with 30 billion parameters running on Apache 2.0, built for persistent local agents. Thanks to 4-bit quantization, the memory requirement drops from 55 GB to 18-20 GB, allowing the model to run on a single consumer GPU. Tech My Money AI Weekly

And here's a crucial point you won't find in most reports: All benchmark figures so far come from the provider itself; no independent evaluation exists, and the training data is not described. Anyone deriving architectural decisions from this should conduct their own measurements. That's precisely what our Technology Due Diligence is for, and AI Sourcing helps answer the question of whether you really want to plan your agent workloads for the next twelve months on a paid hosted API. Warp2Search


💡 Key findings

  • Acceptance is becoming the scarcest resource. In Vancouver, around 600 citizen complaints were enough to postpone a hearing beyond the election. Only one in three Canadians trusts AI.

  • Compute is going local. A 30-billion-parameter model under Apache 2.0 runs on a single graphics card. The make-or-buy threshold for agents shifted this week.

  • Supplier figures are not metrics. No independent evaluation of Muse Glimmer exists to date. Research before you decide, not after.

  • Those who fail to commercialize lose out on added value. Four of Toronto's best chip startups are now owned by US corporations or have relocated there.

  • Infrastructure is the new bottleneck, not the model. Cooling accounts for 30 to 45 percent of a data center's power consumption. Scaling AI means scaling water and network capacity along with it.


📚 Sources


Conclusion

Vancouver reveals both sides of the same equation within just a few blocks. On one side, capital, talent, and a university network that has endured for decades. On the other, 600 people who want to know how much the thing next to their house consumes. Both are real. Anyone who only plans for one side is missing the point.


The companies that win in the next two years are not those with the biggest models. They are the ones that know their own starting point, independently vet their suppliers, and bring their people along.

Want to know where your company truly stands? Start with the AI Readiness Check at readiness.oakai.de or write to us directly at info@oakai.de .

We'll be in touch again from Vancouver ;-)


All services at a glance: oakai.de/services

The future is not a matter of chance. It is a choice.


 
 
 

Comments

OAKAI®

E-Mail: info@oakai.de

NEWSLETTER

Sign up for the OAKAI newsletter.

  • LinkedIn

© 2024 OAKAI®

Imprint

Data protection

Thanks for subscribing!

bottom of page