DeepInfra's Toronto Bet Signals AI's Next Gold Rush: Inference

PALO ALTO, CA – July 08, 2026 – AI inference specialist DeepInfra announced today the opening of its first international data center in Toronto, a move that provides a stark illustration of the AI industry's most significant and capital-intensive pivot to date. The new 1.7 MW facility, which will house over 1,000 of NVIDIA’s latest Blackwell B300 GPUs, is more than just a geographic expansion; it’s a high-stakes bet on where the real value in artificial intelligence will be captured in the coming decade: not in training models, but in running them.

For the past several years, the AI narrative has been dominated by the colossal task of training ever-larger models, a process demanding immense, centralized computing power. Now, the industry is shifting its focus to inference—the operational phase where trained models generate predictions, answer queries, and power applications at scale. This move from the lab to live production is creating an insatiable demand for a new kind of infrastructure: globally distributed, ultra-responsive, and ruthlessly cost-efficient. DeepInfra’s Canadian foray is a critical signal that the infrastructure battle for production AI is officially underway.

The Inference Imperative

The AI gold rush is entering its second act. If the first was defined by building the models, the second is about deploying them into the fabric of daily life and business. This transition is placing unprecedented strain on global compute resources. According to projections from McKinsey & Company, AI inference workloads are on track to constitute over 40% of all data center demand by 2030, expanding at a compound annual growth rate of roughly 35%.

This explosive growth is driven by the mass adoption of real-time applications. From conversational AI assistants and autonomous systems to high-volume API traffic powering enterprise software, the utility of AI is directly tied to its speed. The fractional-second delays, or latency, that were once acceptable in experimental phases are now intolerable in production environments where user experience and operational efficiency are paramount.

“Enterprises are moving from experimentation to production at unprecedented speed, and that shift demands infrastructure that is both scalable and globally distributed,” said Nikola Borisov, CEO and co-founder of DeepInfra, in today's announcement. “This Toronto cluster is a foundational step in expanding our capacity beyond the U.S. and ensuring customers can run AI workloads closer to where their users and data reside.”

Borisov’s statement cuts to the core of the issue. The physics of data transmission means that proximity matters. By placing powerful GPU clusters in key regional hubs like Toronto, companies can slash latency, improve performance, and potentially navigate the increasingly complex web of international data sovereignty regulations.

Toronto's Tech Ascent: A Strategic North American Beachhead

DeepInfra’s choice of Toronto for its ninth data center and first international location is a calculated one, reflecting the city's emergence as a premier global AI hub. The decision transcends mere geographic convenience; it's an investment in a thriving ecosystem built on talent, research, and deliberate government strategy.

Toronto is home to one of North America’s largest and fastest-growing tech talent pools. This workforce is anchored by world-renowned academic institutions like the University of Toronto, a cradle of AI innovation, and the Vector Institute, an independent research organization dedicated to deep learning that attracts top-tier global talent. This concentration of expertise creates a virtuous cycle of innovation and commercialization.

Furthermore, the Canadian government is actively courting this type of investment. The nation recently launched a $2 billion Sovereign AI Compute Strategy, a five-year plan explicitly designed to bolster domestic compute capacity and support the country's AI innovators. With dedicated funds to help small and medium-sized businesses purchase AI compute resources, the government is not just building a field of dreams but ensuring there are teams ready to play on it. DeepInfra's facility plugs directly into this national ambition, providing the very commercial, high-performance infrastructure the strategy aims to foster.

For a company like DeepInfra, this alignment of public policy and private infrastructure creates a powerful synergy. It gains access to a rich talent pool and a supportive policy environment, while Canada gains a crucial piece of the puzzle needed to keep its AI ecosystem competitive on the world stage.

The GPU Gold Rush and the Specialized Cloud

The deployment of over 1,000 NVIDIA Blackwell B300 GPUs is a testament to the immense capital required to compete in the inference market. With individual B300 GPUs estimated to cost upwards of $50,000 and global availability remaining tight, securing such a large allocation is a significant strategic win that follows the company's recent Series B funding round. These chips represent the bleeding edge of AI hardware, promising up to a 30-fold increase in inference performance for large language models compared to the previous generation.

This hardware arms race is forcing a bifurcation in the cloud market. On one side are the hyperscale giants like Amazon Web Services, Microsoft Azure, and Google Cloud, which offer a vast portfolio of services. On the other is a growing cohort of specialized cloud providers—including DeepInfra, CoreWeave, and Together AI—that are purpose-built for AI workloads. These specialists aim to outmaneuver the giants by offering superior performance-per-dollar, deeper expertise, and infrastructure finely tuned for the specific demands of training or inference.

DeepInfra's strategy hinges on high-throughput, cost-efficient inference for the hundreds of open-source models that are gaining traction in the enterprise. By owning and operating its own infrastructure and optimizing its software stack, the company claims it can process trillions of tokens per week at a lower cost than many competitors, a critical advantage for companies deploying high-volume agentic systems or other token-intensive applications.

A Catalyst for Canada's AI Ecosystem

The most immediate impact of DeepInfra's new data center will be felt within Canada's own technology sector. For years, Canadian startups and enterprises have often had to rely on US-based data centers to access top-tier AI infrastructure, incurring higher latency and navigating cross-border data transfer complexities. The arrival of a state-of-the-art inference facility on Canadian soil changes that dynamic.

Local access to Blackwell B300 GPUs can dramatically lower the barrier to entry for Canadian innovators, allowing them to build and deploy sophisticated AI applications with the same low-latency advantages previously reserved for firms located closer to data centers in Virginia or California. This democratization of access is vital for fostering a competitive startup scene and enabling established Canadian companies to integrate AI more deeply into their operations.

Moreover, the facility directly addresses the growing importance of data sovereignty. For industries like finance, healthcare, and the public sector, the ability to process sensitive data within national borders is not just a preference but a regulatory necessity. DeepInfra's Toronto location provides a powerful new option for these organizations. By planting a flag in Toronto, DeepInfra is not just expanding its footprint; it is providing a crucial piece of enabling infrastructure that can accelerate innovation across an entire national ecosystem.