Ontario Puts AI Data Centres Under a Policy Framework, and It Wants Full Price for the Power
The Ontario government announced a Data Centre Playbook on August 14, a framework built on three pillars: economic development, digital sovereignty, and community benefits, with a 30 day public consultation now open. The province plans legislation requiring data centres to pay full price for the electricity they consume, ending an implicit subsidy, while competing for investment on non financial strengths: clean power, available land, climate, and workforce. The announcement lands amid intensifying municipal resistance, with Oakville holding a moratorium on new data centres and Mississauga proposing a pause.
Provinces are now openly diverging on AI infrastructure. Manitoba rejected a major data centre proposal, while Alberta courts development, including the $13 billion hyperscale facility Meta announced near Edmonton in July. Ontario is betting that grid cleanliness and policy clarity will attract the builds that matter, even at full electricity cost.
- Three pillar framework: economic development, digital sovereignty, and community benefits, with a 30 day consultation open
- Planned legislation would require data centres to pay full price for electricity
- Oakville has a moratorium on new data centres and Mississauga has proposed a pause
- Manitoba rejected a major proposal while Alberta courts hyperscale development
Enterprise Impact: Where and how AI capacity gets built in Canada is now an active policy question, and this framework will shape hosting costs, siting timelines, and sovereignty positioning for every enterprise weighing Canadian data residency against available capacity. Full price electricity changes the economics of Ontario hyperscale builds and may push some projects west, but the digital sovereignty pillar could translate into procurement leverage for Canadian hosted AI services. Enterprises with residency mandates should read the consultation documents and respond within the 30 day window, because the definitions this framework settles on will flow into contracts.
Source: BetaKitRBC and BMO Sell Moneris to US Private Equity in a $2 Billion Deal, and One in Three Canadian Transactions Changes Hands
US private equity firm Francisco Partners is acquiring Toronto based Moneris, one of Canada's largest payment processors, from co owners RBC and BMO for $2 billion CAD, BetaKit reported August 11. Moneris supports more than five billion transactions annually and processes roughly one in three Canadian transactions. The deal is expected to close in early 2027 pending regulatory approval, with RBC and BMO splitting proceeds evenly and agreeing to exclusively refer customers to Moneris going forward. Jeff Sloan, the former Global Payments chief executive, becomes chairman under the new ownership.
- Francisco Partners pays $2 billion CAD for the processor behind roughly one in three Canadian transactions
- RBC and BMO split proceeds evenly and will exclusively refer customers to Moneris
- Former Global Payments chief executive Jeff Sloan becomes chairman
- Close is expected in early 2027, subject to regulatory approval
Enterprise Impact: A core piece of Canadian payments infrastructure moves to US private equity ownership, which matters to any enterprise transacting on Moneris rails and to boards tracking foreign ownership of critical Canadian digital infrastructure. Private equity ownership typically brings aggressive platform modernization alongside pricing discipline, so merchants should anticipate investment in the stack and watch fee schedules through the transition. Paired with this week's Council of Canadian Innovators report on scaleups selling abroad, the sovereignty thread is hard to miss.
Source: BetaKitCanadian Tech Firms Sell to Foreign Buyers Exactly When It Is Time to Scale, the CCI Finds
The Council of Canadian Innovators identified what it calls a scale conversion gap: Canadian technology companies exit to international buyers at precisely the moment growth becomes most capital intensive, based on a study of 30 businesses across software, life sciences, energy, finance, and hardware, BetaKit reported August 11. The barriers are consistent: difficulty landing domestic customers, unavailable growth financing, executive talent shortages, and fragmented support programs. Companies often needed international validation before Canadian customers would buy from them.
- 93% of acquired firms kept Canadian operations, but 93% saw leadership and decision making shift abroad
- Domestic customers frequently waited for international validation before buying
- Recommendations include using public procurement to validate emerging firms and building sector expert funds able to lead major rounds
Enterprise Impact: The report puts numbers on Canada's headquarters drain, and its most actionable finding points at buyers, not founders: Canadian enterprises that decline to purchase from domestic scaleups until a foreign acquirer validates them are part of the mechanism. Procurement policy is, in aggregate, industrial policy. Technology leaders who want a deeper Canadian vendor bench for AI, connectivity, and infrastructure have a direct lever in how their own evaluation criteria treat emerging domestic suppliers.
Source: BetaKitConvictional Shuts Down and Returns Half Its Capital After the AI Era Pivot Falls Short
Kitchener Waterloo company Convictional, founded by former Shopify employees, will shut down on August 27 and return slightly less than half of the roughly $49 million US it raised, after its pivot from B2B commerce to an AI era Slack alternative failed to find traction, BetaKit reported August 12. Chief executive Roger Kirkness was direct about the failure mode: larger firms built their own tools internally on foundation models, while smaller companies could not be reached through sustainable distribution. Its earlier product, Modern Dropship, had reached 3,000 customers and $2 million US in annual revenue before being divested in early 2025.
- Roughly $49 million US raised from investors including Y Combinator's growth fund and Garage Capital; just under half returned
- Large enterprises chose to build AI collaboration tooling internally rather than buy
- Shutdown takes effect August 27, 2026
Enterprise Impact: The failure is instructive for buyers as much as builders: in the AI collaboration category, large enterprises increasingly assemble their own tooling directly on model APIs, collapsing the market for standalone AI workspace products. That validates a build calculus for workflow layer AI where the platform does the heavy lifting, and it is a caution when evaluating point solution vendors whose differentiation sits in a thin layer above the model. The disciplined capital return is also a maturing ecosystem norm worth noting.
Source: BetaKitTelesat Posts a $559 Million Loss on Paper While Its Lightspeed Backlog Climbs to $5.6 Billion
Ottawa based satellite operator Telesat reported a $559 million second quarter net loss on August 13, swinging from a $76 million gain a year earlier, with revenue down 25% to $79 million and the stock off 14% on the day. The loss is largely non cash: fair value charges on Lightspeed financing warrants, which surged more than 50% in value after Telesat signed its largest ever contract, plus currency effects on US dollar debt. The backlog for the Lightspeed low Earth orbit constellation now stands at $5.6 billion, and the stock remains up more than 80% year to date.
- The $559 million loss is driven mainly by non cash warrant and currency charges
- Warrant values rose because Lightspeed's prospects improved after the largest contract signing in company history
- Lightspeed backlog reached $5.6 billion; Canada holds 10% and Quebec 1.87% warrant positions from 2024 loan agreements
Enterprise Impact: Behind the headline loss, demand for sovereign Canadian low Earth orbit connectivity is accelerating, and a $5.6 billion backlog matters to enterprises planning connectivity for remote operations, northern infrastructure, and defence adjacent work. The accounting irony, warrants ballooning in value because the business improved, is a reminder to read Canadian technology earnings past the net loss line. Ottawa's warrant positions underline how intertwined the federal government now is with strategic connectivity infrastructure.
Source: BetaKitXanadu Partners With the University of Alberta to Aim Quantum Algorithms at Cancer Therapy Molecules
Toronto quantum company Xanadu and University of Alberta chemistry professor Alex Brown will develop quantum computing algorithms to design next generation photosensitizers, the light reactive molecules used in photodynamic cancer therapy, BetaKit reported August 13. Standard computational methods struggle to model how these molecules behave under light, a simulation task where quantum computers hold theoretical advantage. It is Xanadu's second academic partnership announced this month, following an education memorandum with the University of Guelph, and the company recently published supporting research on simulating interactions between light and matter.
- The partnership targets photosensitizer design for photodynamic cancer therapy
- Molecular behaviour under light is a simulation problem where quantum methods hold theoretical advantage
- Xanadu is now publicly traded, with rising research spending and expanded US operations in Albany, New York
Enterprise Impact: Quantum utility is arriving first through narrow, high value simulation problems in chemistry and materials, not general computing, and Canadian institutions are visibly in that first wave. For technology strategists, this is the pattern to track for quantum readiness planning: watch domain partnerships, not qubit counts. Xanadu's public listing and expansion keep Canada credibly positioned in a second strategic technology alongside AI.
Source: BetaKitNVIDIA Enlists Six Asset Managers to Mobilize More Than $500 Billion for AI Compute, and Compute Becomes an Asset Class
NVIDIA signed memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR on August 10 to establish independent financing platforms intended to mobilize more than $500 billion US of third party capital for AI compute infrastructure over time. The platforms will offer dedicated pools of capital at attractive rates to NVIDIA customers building AI factories, without the buildout sitting on NVIDIA's balance sheet. Chief executive Jensen Huang framed the thesis in five words: in AI, compute is revenue.
- Six partners: Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR, structured as memorandums pending final agreements
- Target is more than $500 billion of third party capital, treating compute like infrastructure assets such as toll roads and real estate
- NVIDIA argues software updates extend the useful life of the hardware being financed
Enterprise Impact: This formalizes AI compute as a distinct capital asset class, which changes how large scale capacity gets funded and priced. Deep institutional capital behind GPU infrastructure should ease capacity constraints and stabilize long term compute pricing, but it also means multi year compute commitments are becoming financial instruments, so procurement teams should expect more leasing, financing, and capacity contract structures in vendor conversations. Brookfield's seat at the table puts Canadian institutional capital directly into the global AI buildout.
Source: NVIDIAAnthropic's IPO Math Comes Into View: A $190 to $200 Billion Revenue Forecast for 2028
As Anthropic prepares one of the largest initial public offerings on record, Reuters reported August 14 that the company is projecting roughly $190 billion to $200 billion in 2028 revenue, the figure bankers are using to price the listing on forward enterprise value to revenue multiples. Bloomberg separately reported that Anthropic is telling prospective investors second quarter revenue jumped at least 14x year over year, with run rate growing from about $9 billion at the end of 2025 to more than $47 billion by May 2026 and a first quarterly operating profit of $559 million anticipated.
- 2028 revenue projected at roughly $190 billion to $200 billion, per sources cited by Reuters
- Second quarter 2026 revenue projected at $10.9 billion or more, at least 14x year over year
- Valuation comparables in circulation include Palantir at 53x expected revenue and SpaceX and Cloudflare at about 41.6x
Enterprise Impact: The AI model layer is about to receive public market discipline, and the numbers being underwritten assume enterprise AI spending grows several fold within two years. For technology leaders this cuts two ways: a public Anthropic means more financial transparency and durability signal for a strategic vendor, while valuations priced on 2028 forecasts imply sustained pricing power in enterprise AI contracts. Vendor risk assessments for foundation model providers should now include post IPO behaviour on pricing, roadmap, and enterprise support.
Source: BloombergOpenAI's Specialized Security Models Land on Amazon Bedrock, Behind an Eligibility Gate
OpenAI and AWS made the Daybreak Red and Daybreak Blue security models available on Amazon Bedrock on August 11, restricted to customers enrolled in OpenAI's Trusted Access for Cyber program. Daybreak Blue provides GPT 5.6 Sol with defensive security safeguards for detection engineering and incident response; Daybreak Red offers GPT 5.6 Cyber, purpose trained for vulnerability research and exploit analysis. The models run on Bedrock's inference engine with zero operator access at the chip level, inference data excluded from training, and zero data retention available on request.
- Access is gated through OpenAI's Trusted Access for Cyber program, initially in the US East Ohio region
- Use cases span vulnerability discovery, detection engineering, incident response, and source code analysis at scale
- Controls include zero operator access, no training on inference data, and optional zero data retention
Enterprise Impact: Frontier AI for security work can now run inside existing AWS governance boundaries rather than through a separate OpenAI relationship, removing a major procurement and data control objection. The gated access model, eligibility programs wrapped around dual use capability, is a template other vendors will copy, and worth understanding now because it changes how these capabilities are bought. The deepening OpenAI and AWS relationship, three months after OpenAI models first reached Bedrock, also reshuffles assumptions about model exclusivity on hyperscale clouds.
Source: AWSAnthropic Will Watermark AI Generated Text, and Provenance Marking Becomes Default Platform Behaviour
Anthropic announced it will add watermarks to text generated by its models to comply with European transparency regulations, with marks designed to travel with copied text and potentially persist through editing, BetaKit reported August 12. Google, Meta, Microsoft, and OpenAI have made similar commitments to identify AI generated content. The same coverage notes the uncomfortable research consensus: AI detection tools remain unreliable, which is why institutions are shifting to clear usage policies with human accountability rather than detection tooling.
- Watermarks persist through copy and paste and may survive editing; exact mechanics were not specified
- All major model providers have now committed to content identifiers, driven by European requirements
- Research cited shows detection tools do not work reliably, pushing organizations toward usage policies instead
Enterprise Impact: Provenance marking of AI output is becoming a default platform behaviour rather than an opt in, and enterprises need to map where watermarks will surface in their content pipelines before client deliverables, marketing copy, and code carry machine readable AI attribution. The practical governance answer is the one the coverage points to: a clear internal AI usage policy with named human accountability, not detection tools that the research says cannot be trusted. Review vendor terms now to understand what marking applies to output your organization publishes.
Source: BetaKitIBM Builds a Dedicated OpenAI Practice, and Enterprise AI Competition Shifts From Models to Distribution
IBM and OpenAI announced a strategic alliance on August 13 covering joint marketing and industry specific solutions for financial services, government, telecommunications, and retail. IBM will build a dedicated OpenAI practice inside IBM Consulting and train tens of thousands of consultants on GPT 5.6, Codex, and ChatGPT Work, with OpenAI models integrated into IBM Consulting Advantage and IBM Autonomous Security. The alliance arrives less than a year after IBM struck a similar arrangement with Anthropic, and IBM remains model agnostic through watsonx and its own Granite models.
- A dedicated OpenAI practice inside IBM Consulting, with tens of thousands of consultants trained
- Target industries are financial services, government, telecommunications, and retail
- IBM now holds parallel alliances with both OpenAI and Anthropic while staying model agnostic
Enterprise Impact: The battle for enterprise AI spend is shifting from model quality to distribution, and OpenAI just gained a services channel into the largest regulated enterprises through IBM's consulting bench. The systems integrator you already work with increasingly determines which model families arrive in your stack, which makes model selection a partner governance question as much as a technical one. IBM's dual alliances also validate the multi model strategy as the enterprise default rather than the exception.
Source: TechCrunchThrive Holdings Raises $2 Billion at a $12 Billion Valuation to Buy Traditional Businesses and Rebuild Them on AI
Thrive Holdings, the Thrive Capital spinout that operates like a private equity firm for AI, raised $2 billion from SoftBank, D1 Capital Partners, and Altimeter Capital at a $12 billion valuation, TechCrunch reported August 12. The firm buys traditional businesses such as accounting and IT services companies and embeds AI in their operations, with OpenAI holding an ownership stake since December 2025. It already operates more than 70 businesses across two platforms and will use the new capital to launch a third focused on regulatory services for physical infrastructure: data centres, manufacturing, healthcare, power, water, and transportation.
- Existing platforms: Current, spanning more than 50 accounting firms, and Shield, covering roughly 20 IT companies
- Claimed results include 7,000 plus tax returns processed at 98% accuracy and help desk resolution accelerated 36x
- The third platform targets regulatory and permitting services for physical infrastructure
Enterprise Impact: The AI rollup thesis, buying services businesses and rebasing their economics on AI, just received a $12 billion endorsement, and OpenAI's equity stake shows model providers moving directly into services economics. For executives, this is a preview of AI native competitors in professional services: the same services at structurally lower cost. The new infrastructure compliance vertical is its own signal, marking permitting and certification workflows as the next category judged ready for AI automation at scale.
Source: TechCrunchGoogle's Pixel 11 Puts a 2nm Chip and Autonomous Agents in Pockets, and On Device AI Gets Serious
Google announced the Pixel 11 lineup on August 12 at Made by Google 2026, built on the Tensor G6, the first 2nm chip in an Android phone. The new tensor processing unit delivers 50% more compute and double the memory bandwidth of its predecessor, running on device AI up to 3.5x faster at up to 3.5x less energy. Gemini gains agentic abilities on the device, including autonomously ordering groceries and booking rides across third party apps, and Live Transcribe adds support for American Sign Language. The Pixel 11 starts at $899 US with most of the lineup shipping August 20.
- Tensor G6 is the first 2nm chip in an Android phone, with 50% more compute and double the memory bandwidth
- On device AI runs up to 3.5x faster at up to 3.5x less energy than the prior generation
- Gemini can act autonomously across third party apps; accessibility gains include sign language transcription
Enterprise Impact: The efficiency jump in on device silicon matters more to enterprises than the phones: fast, low power local inference shifts AI workload off the cloud, with implications for data governance, latency sensitive field applications, and mobile fleet strategy. Agentic features acting across apps preview the permission and audit questions IT will face as autonomous agents reach managed devices. Device refresh conversations should start including neural processing capability as a selection criterion, not an afterthought.
Source: TechCrunch