AMD Moves to Acquire Toronto's Taalas, and Canada's Most Advanced AI Chip Startup Heads South
AMD is acquiring Taalas, the Toronto semiconductor startup that converts AI models into custom silicon for inference, BetaKit reported August 6. Founded in 2023 by Ljubisa Bajic, Lejla Bajic, and Drago Ignjatovic, Taalas hardwires a specific model into a chip and claims it can turn a new design around in roughly two months against an industry norm of one to two years. The company had raised $219 million US across two rounds: $50 million in 2024 led by Quiet Capital and Pierre Lamond, and $169 million earlier this year with participation from Fidelity. Deal terms were not disclosed. AMD senior vice president Vamsi Boppana said the acquisition will "strengthen our AI portfolio by delivering differentiated inference performance and efficiency."
It is AMD's second acquisition of a Canadian AI chip team in roughly a year, following the Untether AI engineering team in 2025, and chief executive Ljubisa Bajic previously co founded Tenstorrent, the other flagship of Canadian AI silicon. The pattern is consistent: Canada produces frontier chip design talent, and the exits keep consolidating that talent into American acquirers, this time two months after Ottawa published an AI strategy built around sovereign capability.
- Taalas hardwires AI models into custom inference silicon and claims a roughly two month turnaround for a new chip
- The company raised $219 million US across 2024 and 2026 rounds from backers including Quiet Capital, Pierre Lamond, and Fidelity
- This is AMD's second Canadian AI chip acquisition in about a year, after the Untether AI team in 2025
- Deal value was not disclosed; the announcement came August 6
Enterprise Impact: Inference, not training, is becoming the dominant line in enterprise AI budgets, and AMD just bought a shortcut to model specific silicon aimed exactly at that cost curve. The near term signal for technology leaders is that credible alternatives to general purpose GPUs are arriving faster than most roadmaps assume, which strengthens the buyer's hand in 2027 capacity planning and makes inference efficiency a fair question to put to every AI platform vendor. For Canadian executives building a sovereign AI procurement case, the deal is also a caution: the domestic chip design bench keeps thinning at exactly the layer sovereignty arguments depend on.
Source: BetaKitShopify's Quarter Makes Agentic Commerce a Revenue Line: Orders From AI Powered Searches Up Nearly 13x
Shopify reported second quarter revenue of $3.58 billion US on August 5, up 34% year over year and ahead of estimates, with net income of $1.5 billion and gross merchandise volume of $115 billion, up from $87.8 billion a year earlier. Shares rose about 17% on the Toronto Stock Exchange to around $203 CAD. The number that matters for technology strategy: orders originating from AI powered searches grew nearly 13x, and 75% of those AI attributed orders came from outside Shopify's top 100 categories. President Harley Finkelstein called Shopify "probably the most AI pilled company in the world." Third quarter guidance puts revenue growth in the low 30% range.
- Revenue of $3.58 billion US, up 34%, with gross merchandise volume of $115 billion, up from $87.8 billion a year earlier
- Orders from AI powered searches grew nearly 13x, with 75% of AI attributed orders outside the top 100 categories
- Net income reached $1.5 billion and shares rose about 17% on the results
- Third quarter guidance holds revenue growth in the low 30% range
Enterprise Impact: This is the clearest Canadian evidence yet that agentic commerce is a functioning revenue channel rather than a demo. A 13x jump in AI originated orders, concentrated in long tail categories, means discovery is shifting toward AI agents that read structured catalogs rather than shoppers who browse. Retail and consumer facing enterprises should treat agent readiness, meaning structured product data, clean APIs, and checkout an agent can complete, as a 2026 to 2027 roadmap item with measurable share implications, not an innovation experiment.
Source: BetaKitSanja Fidler, Who Built NVIDIA's Toronto Lab From One Researcher to More Than Eighty, Steps Down as VP of AI Research
Sanja Fidler is leaving NVIDIA after eight years as vice president of AI research, BetaKit reported August 4. Fidler founded and led NVIDIA's Toronto Spatial Intelligence Lab, growing it from employee number one to more than 80 researchers, with milestones including real time interactive driving in AI generated environments. A co founder of the Vector Institute, she remains an associate professor at the University of Toronto. Her next destination is undisclosed; she says world models are where the next breakthrough lies.
- Fidler led NVIDIA's Toronto Spatial Intelligence Lab for eight years, growing it to more than 80 researchers
- She co founded the Vector Institute and remains an associate professor at the University of Toronto
- Her next role is undisclosed; she points to world models as the site of the next breakthrough
- The departure lands the same week AMD announced its acquisition of Taalas
Enterprise Impact: The departure of one of Canada's most senior industrial AI researchers, in the same week the country's leading AI chip startup agreed to sell to a US acquirer, sharpens the question of whether Canada retains the leadership layer of its AI ecosystem, not just the graduates. Her bet on world models is also a useful directional signal for enterprise roadmaps: simulation grade AI for physical environments is where frontier research talent is moving, which is worth weighing in digital twin, robotics, and industrial automation plans.
Source: BetaKitOttawa Backs Canada's First Quantum Defence Innovation Hub With More Than $20 Million, Led by the University of Calgary
The federal government is investing more than $20 million over two years in a 13 member consortium led by the University of Calgary to create Canada's first Defence Innovation Secure Hub, focused exclusively on quantum technology, BetaKit reported August 6. Partners include the University of Alberta, University of Lethbridge, University of Saskatchewan, Lockheed Martin Canada, General Dynamics Mission Systems Canada, CAE, Dell Canada, and Calian. The hub builds on Quantum City, the quantum hub the University of Calgary opened in 2025. Defence Minister David McGuinty said quantum will shape the next generation of defence capabilities and that Canada must be prepared to lead rather than follow.
- More than $20 million federal investment over two years in a 13 member consortium
- The Defence Innovation Secure Hub model debuts with an exclusive quantum focus
- Industry partners include Lockheed Martin Canada, General Dynamics Mission Systems Canada, CAE, Dell Canada, and Calian
- The hub extends the Quantum City initiative the University of Calgary launched in 2025
Enterprise Impact: Ottawa is now funding quantum through a defence channel, and defence procurement has a track record of pulling dual use technology toward commercial maturity faster than research grants alone. Enterprises with long lived cryptographic assets should watch hub outputs as an early signal of where post quantum and quantum sensing procurement will land, and Canadian technology firms eyeing the defence supply chain gain a named entry point with both primes and universities at the table.
Source: BetaKitWhite Star Capital Closes a $350 Million Fund With Physical AI at the Centre of Its Thesis
Montreal founded White Star Capital closed its fourth global fund at $350 million CAD after an 18 month raise, BetaKit reported August 5. The fund writes cheques of $5 million to $15 million US at Series A and B, reserves roughly half the fund for follow on investment, and targets physical AI, AI powered commerce infrastructure, healthcare, and financial services. Eight companies are already backed, including Montreal based Tetrix. Canadian limited partners include Fonds de solidarite FTQ, Investissement Quebec, Desjardins, Teralys Capital, and several Canadian banks.
- Fund IV closed at $350 million CAD with cheques of $5 million to $15 million US at Series A and B
- Roughly half the fund is reserved for follow on investment
- The thesis centres on physical AI, AI powered commerce infrastructure, healthcare, and financial services
- Canadian institutional backers include Fonds de solidarite FTQ, Investissement Quebec, Desjardins, and Teralys Capital
Enterprise Impact: A closed $350 million fund with heavy Quebec institutional backing is a liquidity signal in a tight Canadian venture market, and its physical AI thesis previews where enterprise pilot dealflow will come from over the next two years. Corporate innovation and venture client teams should expect a stronger pipeline of Canadian Series A and B vendors in industrial AI, robotics adjacent tooling, and commerce infrastructure, which is a useful counterweight to sourcing every emerging capability from US platforms.
Source: BetaKitAnthropic Puts the Customer's Security Stack in the Inference Path With Hooks for Claude Enterprise
Anthropic launched inference hooks in beta for Claude Enterprise on August 5, a control that routes every governed prompt and tool response through the customer's own security server for an allow or deny verdict before Claude processes it. Coverage spans chat, Claude Code, Claude Cowork, and tool calls including MCP connectors, skills, and plugins. The mechanism works with existing data loss prevention infrastructure from Netskope, Palo Alto Networks, Proofpoint, and Zscaler, or with custom servers built against a published webhook schema, and is configured once at the organization level.
- Every governed prompt and tool response gets an allow or deny verdict from the customer's security server before processing
- Coverage includes chat, Claude Code, Claude Cowork, MCP connectors, skills, and plugins
- Works with existing DLP stacks from Netskope, Palo Alto Networks, Proofpoint, and Zscaler, or custom servers
- Available in beta for Claude Enterprise customers, configured at the organization level
Enterprise Impact: This closes one of the most common objections security teams raise against enterprise AI assistants: the inability to inspect what flows to the model in real time. Extending existing DLP policy enforcement into AI chat and coding sessions through one server side control materially lowers the compliance cost of scaling assistants beyond pilot groups. Expect pressure on Microsoft, Google, and OpenAI to match customer controlled pre inference inspection, and treat its presence or absence as an evaluation criterion in enterprise AI platform selection.
Source: AnthropicOpenAI Removes Text Chat Limits for Free Users and Makes GPT 5.6 Luna the Default
OpenAI announced August 6 that text chats in ChatGPT are becoming unlimited on every tier, with limits remaining only on files, images, and voice. GPT 5.6 Luna replaces GPT 5.5 as the default model for free and Go users, while Plus and Pro subscribers get an upgraded GPT 5.6 Sol. A new Think button appears on all tiers for harder questions, paid users gain an adjustable reasoning slider, and OpenAI says factual errors are 62% less common on Luna and 68% less common on Sol than the models they replace. ChatGPT recently passed 1 billion weekly users.
- Text chat limits removed for all tiers, rolling out this week
- GPT 5.6 Luna becomes the free and Go default; Plus and Pro get GPT 5.6 Sol
- OpenAI cites factual errors 62% less common on Luna and 68% less common on Sol
- ChatGPT recently passed 1 billion weekly users; file, image, and voice limits remain
Enterprise Impact: Free, unlimited, better grounded ChatGPT for a billion weekly users guarantees more unsanctioned AI use inside every organization, which raises the bar for IT to provide sanctioned tools at least as capable, with governance attached. The reasoning slider also normalizes cost and quality tiering per query, a pattern worth mirroring in enterprise AI gateway and cost governance designs rather than routing every request to the most expensive model by default.
Source: TechCrunchAmazon Becomes the Fifth Company to Cross $3 Trillion as the Market Pays for Cloud AI Growth
Amazon's market value closed above $3 trillion for the first time on August 3, making it the fifth company to reach the milestone after Apple, Microsoft, Alphabet, and NVIDIA. Shares hit a record $286.20, up 5.5% on the day, extending the roughly 15% surge of the prior Friday that followed the strongest AWS growth in more than four years: $42.2 billion in quarterly revenue, up about 37%. Amazon added its third trillion in just over two years, having crossed $2 trillion in June 2024.
- Market capitalization topped $3 trillion on August 3, with shares at a record $286.20
- Amazon is the fifth member of the $3 trillion club, after Apple, Microsoft, Alphabet, and NVIDIA
- The move extends a roughly 15% surge driven by the fastest AWS growth in more than four years
- The third trillion in market value arrived in just over two years
Enterprise Impact: Markets are now paying directly for demonstrated AI monetization through cloud, and the AWS reacceleration confirms enterprise AI workloads are landing on hyperscale platforms at production scale. For technology leaders in the middle of contract cycles, hyperscaler pricing leverage is strengthening with every quarter like this one, which makes multi cloud optionality and disciplined committed spend positions more valuable going into 2027 renewals, not less.
Source: CNBCPalantir Grows 93% and Raises Full Year Guidance by Roughly a Billion Dollars as Controlled AI Deployment Sells
Palantir reported second quarter revenue of $1.94 billion on August 3, up 93% year over year against expectations of $1.801 billion, with earnings per share of $0.41 versus the $0.35 expected. US commercial revenue grew 149% and government revenue grew 90%. The company raised full year 2026 guidance to a range of $8.150 billion to $8.158 billion from roughly $7.19 billion, and shares rose 13% in after hours trading. Chief executive Alex Karp told investors the business is compounding at a rate and scale the company has never before witnessed, with management crediting demand for AI sovereignty and controlled AI deployments.
- Revenue of $1.94 billion, up 93%, with US commercial revenue up 149% and government up 90%
- Full year guidance raised by roughly a billion dollars to $8.150 billion to $8.158 billion
- Earnings per share of $0.41 beat the $0.35 expected; shares rose 13% after hours
- Management attributes demand to AI sovereignty and controlled AI deployment requirements
Enterprise Impact: A 149% jump in US commercial revenue is the strongest quarterly evidence yet that large enterprises are paying for operational AI platforms rather than experiments, and that sovereignty, meaning control over where and how AI runs and what it touches, is now a purchasable feature commanding premium growth. The pattern worth noting for platform strategy: budgets are consolidating toward offerings that combine data integration, governance, and AI in one controlled deployment, which raises the bar for internally assembled stacks to prove they can match that integration.
Source: FortuneThe EU AI Act's Transparency Rules Went Live August 2, and Enforcement Now Has an Address
From August 2, providers and deployers of AI systems reaching European Union users must label AI generated content, including deepfakes, and disclose when people are interacting with chatbots, agents, or avatars rather than humans. Enforcement is now active through national market surveillance authorities, the European AI Office, which holds powers over general purpose AI models, and the European Data Protection Supervisor, with fines up to 15 million euros or 3% of global annual turnover and proportionality provisions for smaller companies. The obligations that actually bit this week are transparency and general purpose model supervision: the Digital Omnibus package adopted July 24 deferred high risk system obligations to December 2027 and August 2028.
- AI generated content labelling and chatbot disclosure duties became enforceable August 2
- The European AI Office now supervises general purpose AI model providers directly
- Fines reach 15 million euros or 3% of global annual turnover, with proportionality for SMEs
- High risk system obligations were deferred to December 2027 and August 2028 by the July Digital Omnibus
Enterprise Impact: Any enterprise whose AI outputs reach EU users now carries hard labelling and disclosure duties with real fine exposure, regardless of where the company sits. The near term work is operational rather than legal: inventory customer facing AI touchpoints, add disclosure to conversational interfaces, and put machine readable labelling into generated content pipelines. The high risk deferral buys time on conformity assessments, but transparency compliance is due now, and the enforcement bodies taking complaints include ones competitors can petition.
Source: European Commission