University of Toronto Partners With Cohere to Build an Enterprise Wide AI Platform on Privately Deployed Canadian Technology
On July 23, the University of Toronto announced a partnership with Cohere that places the Toronto AI company's secure, privately deployable agentic platform, North, at the centre of the university's forthcoming enterprise wide AI platform. North will support the coordination of complex tasks across different systems inside one of Canada's largest institutions, with the university framing the partnership as setting the stage for responsible AI adoption at scale. The deployment model matters as much as the deal: North runs privately within the institution's own environment, keeping data inside the organization rather than routing it through a public AI service.
The partnership pairs a Canadian institution serving hundreds of thousands of students, staff, and researchers with a Canadian AI company whose enterprise platform is built for regulated and privacy sensitive environments. For the broader market, it is a template: agentic AI deployed institution wide, on domestic technology, under institutional control.
- The University of Toronto announced the Cohere partnership on July 23
- Cohere's North platform will coordinate complex tasks across systems within a new enterprise wide AI platform
- North deploys privately, keeping institutional data inside the organization's own environment
- The university positions the partnership as a foundation for responsible AI adoption at scale
Enterprise Impact: This is one of the most concrete Canadian examples yet of agentic AI procured at institutional scale, and it validates private deployment as the pattern for organizations that cannot send sensitive data to public AI services. Technology leaders in regulated sectors should study the structure: a single enterprise platform rather than scattered pilots, a privately deployed agentic layer, and governance framed up front. It also strengthens the commercial case for Canadian sovereign AI options when data residency, privacy obligations, or institutional trust are at stake. Organizations planning enterprise AI platforms should shortlist deployment models, not only model capability, when evaluating vendors.
Source: University of TorontoClio Brings Its AI Legal Workspace to Canada, Grounded in a Dataset of More Than 470,000 Canadian Cases
Burnaby headquartered Clio launched Clio Work in Canada on July 23, bringing its AI workspace for legal professionals to the domestic market after earlier availability in the United States and select markets. The product offers case summarization, case law research, legal argument suggestions, and document drafting, operating within a closed system grounded in verifiable Canadian case law: a dataset spanning more than 470,000 cases across more than 40 courts, assembled through Clio's acquisition of Toronto based Jurisage. The company says Clio Work has been its fastest adopted product ever, and the Canadian launch follows the $1.4 billion CAD acquisition of vLex that enabled its United States expansion.
- Clio Work launched across Canada on July 23 after earlier availability in the United States
- The system is grounded in more than 470,000 Canadian cases across more than 40 courts
- The Canadian dataset came through Clio's acquisition of Toronto based Jurisage
- Clio describes Clio Work as the fastest adopted product in its history
Enterprise Impact: The launch shows what defensible vertical AI looks like: proprietary, verifiable domain data fencing a closed system, rather than a thin layer over a general purpose model. Enterprises evaluating sector specific AI tools should weigh the provenance and verifiability of the underlying data as heavily as the model features, since grounded systems materially reduce fabrication risk in professional work. For Canadian firms, domestic grounding also localizes the accuracy question: a legal AI trained on Canadian authority is answering the question Canadian professionals actually ask. Expect this playbook, buying the dataset that grounds the product, to repeat across other regulated verticals.
Source: BetaKitMaluuba's Co Founders Launch Skyfall AI, Betting That World Models Can Run Entire Businesses Autonomously
Sam Pasupalak, Sumit Pasupalak, and Kaheer Suleman, whose deep learning startup Maluuba was acquired by Microsoft in 2017, launched Skyfall AI on July 24 with backing from Inovia Capital and Garage Capital. The company builds AI world models, systems trained on physical and spatial data rather than text alone, which proponents argue reason about cause and effect rather than matching patterns, making them better suited to long horizon planning and high stakes decisions under uncertainty. Skyfall's opening demonstration is deliberately provocative: acquire a small B2B software business for up to US$1 million, install an AI chief executive, and attempt to double the company's revenue within six months.
- Skyfall AI launched July 24, founded by the former Maluuba co founders and backed by Inovia Capital and Garage Capital
- World models train on physical and spatial data to build representations that support cause and effect reasoning
- The company plans to acquire a small B2B software business and run it autonomously with an AI chief executive
- The stated goal is to double the acquired company's revenue within six months
Enterprise Impact: The claim deserves skepticism until the demonstration lands, but the direction matters: a credible Canadian research team is betting that the next capability frontier is autonomous operation of business processes end to end, not better chat. Technology leaders should track world models as a distinct research track from large language models, because success would change the automation conversation from tasks to functions. In the nearer term, the launch is another signal of Canadian AI research depth converting into companies, and a reminder that governance frameworks for AI autonomy will be tested by ventures explicitly designed to remove humans from the loop.
Source: BetaKitMicrosoft Commits Azure to AMD's Helios Rack Scale Systems, Opening a Second Front in AI Infrastructure
AMD announced an expanded strategic partnership with Microsoft on July 20 under which Azure will deploy the AMD Helios rack scale platform to power frontier model inference for Microsoft, its AI customers, and Azure AI services. Helios integrates 72 Instinct MI455X GPUs, sixth generation EPYC Venice processors, Pensando networking, and the ROCm software stack into a single open standard rack, with shipments to customers including Microsoft beginning in the second half of 2026. Azure will also add two new virtual machine families on EPYC Venice processors and deploy Pensando data processing units in its AI backend networking.
- Azure will deploy AMD Helios racks for frontier model inference across Microsoft and customer workloads
- Each Helios rack integrates 72 Instinct MI455X GPUs with EPYC Venice CPUs and Pensando networking
- Shipments begin in the second half of 2026; two new EPYC Venice VM families are also coming to Azure
- The deal is one of AMD's most significant cloud design wins to date
Enterprise Impact: A credible second source for rack scale AI infrastructure matters to every enterprise buying AI capacity, because single vendor supply has been a structural driver of cost and scarcity. If Helios lands at scale on Azure, inference pricing pressure follows, and enterprises negotiating multi year AI platform commitments should factor that trajectory into timing and terms. The open standard rack design also hints at a future where AI infrastructure procurement looks more like traditional compute buying, with competing integrated platforms, than like the allocation queue of the past three years. Watch second half availability before treating the competition as real capacity.
Source: AMDAlphabet and Tesla Beat on Earnings but Lose Hundreds of Billions in Value as AI Capital Spending Shocks the Market
Second quarter earnings from Alphabet and Tesla, reported July 22, delivered the first real test of AI capital spending at scale, and the market's answer was severe. Alphabet beat expectations, with Google Cloud revenue up 82% to $24.8 billion, yet raised its full year capital expenditure outlook to between $195 billion and $205 billion, more than double last year, after spending $44.9 billion in the quarter alone. Tesla's capital spending rose 142% to $5.79 billion with roughly $25 billion guided for the year, pushing free cash flow negative by $1.1 billion. Tesla shares fell 12% and Alphabet 6% on July 23, erasing hundreds of billions in combined market value.
- Alphabet raised full year capital spending guidance to between $195 billion and $205 billion, more than double last year
- Google Cloud revenue grew 82% to $24.8 billion in the quarter
- Tesla capital spending rose 142% to $5.79 billion and free cash flow turned negative by $1.1 billion
- Both companies beat earnings expectations; the selloff was a response to spending guidance
Enterprise Impact: The market is now pricing discipline into AI infrastructure, and that discipline will flow downstream. Enterprises should expect cloud providers under investor pressure to push harder on committed spend agreements, reserved capacity, and premium pricing for scarce accelerators, while also becoming more willing to negotiate as they seek revenue to justify the buildout. The 82% growth in Google Cloud is the counterweight: enterprise AI demand is real and accelerating. Technology leaders should use this moment of investor scrutiny to negotiate multi year terms, because providers need demonstrated demand precisely when the market is questioning their spending.
Source: CNBCGartner Forecasts the AI Platforms and Models Market Will Grow 63% to $64 Billion in 2026, With Governance Vendors the Biggest Winners
Gartner forecast on July 20 that worldwide end user spending on AI models and platforms will reach $64 billion in 2026, up 63.4% from $39 billion in 2025. Within that, spending on generative AI models is forecast to grow 117% while AI platform spending rises 36.9%. Gartner's analysts note that enterprise AI budgets are coming under greater scrutiny, with increased focus on usage efficiency, cost control, and measurable outcomes, and argue the biggest winners will be vendors that help enterprises manage where and how AI is used, embedding evaluation, cost transparency, and usage tracking into customer workflows.
- Worldwide AI platforms and models spending is forecast at $64 billion in 2026, up 63.4% year over year
- Generative AI model spending is forecast to grow 117%; platform spending grows 36.9%
- Gartner flags rising scrutiny of AI budgets and demand for measurable outcomes
- Vendors that help enterprises govern usage and cost are positioned to win
Enterprise Impact: The forecast quantifies a shift from experimentation budgets to production spending, and it validates governance as a market force rather than a compliance afterthought: the winning vendors are the ones that make AI usage visible, measurable, and controllable. Enterprises should get ahead of that curve internally, standing up usage tracking, cost allocation, and evaluation criteria before spending doubles again, because organizations that cannot measure AI value will struggle to defend AI budgets under the scrutiny Gartner describes. Buying decisions made in 2026 should weight cost transparency and evaluation tooling, not only raw capability.
Source: GartnerIndustry Letter Urging Washington to Protect Open Weight AI Models Doubles to Fifty Signatories in Days, With Amazon and Anthropic Notably Absent
A letter titled Open Weights and American AI Leadership, published July 24 and championed by NVIDIA chief executive Jensen Huang, urged Washington not to restrict open weight AI models as policymakers weigh limits aimed at Chinese open models. The letter launched with roughly 25 signatories including NVIDIA, Microsoft, Meta, IBM, Dell Technologies, Palantir, Hugging Face, Mistral, Andreessen Horowitz, Y Combinator, Mozilla, and the Linux Foundation, and roughly doubled to 50 names within two days as OpenAI and Google joined. Amazon and Anthropic remained absent as of July 26, a conspicuous gap given Amazon's position as Anthropic's largest investor.
- The open weights letter published July 24 with about 25 signatories and reached 50 within two days
- Signatories include NVIDIA, Microsoft, Meta, OpenAI, Google, IBM, Dell, Hugging Face, and the Linux Foundation
- Amazon and Anthropic had not signed as of July 26
- The letter responds to proposals in Washington to restrict open weight models, including Chinese models
Enterprise Impact: Open weight models underpin a large share of enterprise AI deployments, particularly private and on premises installations where data cannot leave the organization. Restrictions on open models, or uncertainty about their future, would directly affect architecture choices that enterprises are making now. Technology leaders relying on open weight models should track this policy debate as a supply chain risk: inventory where open models sit in the stack, note their origin, and ensure swap paths exist if specific model families become restricted. The split among major vendors also previews divergent commercial strategies, worth understanding before committing to any single provider's ecosystem.
Source: Tom's Hardware