Something changed in the AI market this week. Three frontier models launched inside 72 hours — Grok 4.5, OpenAI’s GPT-5.6 family, and Meta’s first paid model — and not one of them led with a benchmark. They led with price. Meanwhile Meta committed more than $50 billion to a single data centre, a major study found that the heaviest AI adopters are hiring more people rather than fewer, and the money kept moving toward the parts of the stack everyone else depends on: chips and human judgment.
1. Three frontier AI models launched in 72 hours — all of them competing on price
SpaceXAI’s Grok 4.5, OpenAI’s GPT-5.6 family, and Meta’s Muse Spark 1.1 all shipped within three days, and every launch led with the same argument: cost, not capability. For any business buying AI, the bill is finally moving in your favour — OpenAI’s cheapest new tier runs $1 per million input tokens against $10 for a premium coding model, so the leverage now sits with buyers who can switch.
2. Meta starts charging for its models and undercuts everyone in the market
Meta released Muse Spark 1.1, its first-ever paid model, priced at roughly a quarter of what OpenAI and Anthropic charge for their top tiers, with Mark Zuckerberg saying rival labs’ pricing is “very extreme and has very high margins.” The company that used to give its models away is now a vendor — and its stock had its best week since early 2024 on the news.
3. OpenAI’s GPT-5.6 goes wide after Washington lifts its restrictions
OpenAI moved GPT-5.6 to general availability with three tiers — Sol, Terra, and Luna — after weeks of government-gated access, and Sam Altman led the pitch with efficiency: Sol uses 54% fewer tokens on agentic coding tasks. The most capable American models are back on the open market, and the sales pitch has shifted from what they can do to what they cost to run.
4. Meta commits more than $50 billion to a single Louisiana data centre
Meta confirmed its Hyperion supercluster in Richland Parish, Louisiana will be a 5-gigawatt facility costing north of $50 billion, backed by generous state tax incentives. The scale of a single site shows how capital-intensive frontier AI has become — and why the labs are now under real pressure to charge for it.
5. OpenAI launches ChatGPT Work, an agent that stays on a project for hours
OpenAI introduced ChatGPT Work, an agent that takes action across a company’s apps and files and can stick with a single goal for hours until it produces finished work. It is the clearest sign yet that the product being sold has moved from a chat window to a coworker that completes tasks.
6. The companies adopting AI hardest are hiring more people, not fewer
A study of 21,559 US firms by Ramp and Revelio Labs found that the heaviest AI adopters grew headcount 10.2% over two years, outpacing companies that adopted little or none. The finding cuts against the prevailing layoff narrative and suggests AI investment has, so far, tracked with growth rather than replacement.
7. Thomson Reuters cuts up to 500 engineering jobs as AI adoption deepens
The Canadian parent of Reuters is cutting as many as 500 engineering roles, citing deeper AI adoption across its technology organization. Set against the research showing AI adopters growing headcount, it captures the real picture: AI is not shrinking companies so much as reshaping which roles they need.
8. SambaNova raises $1 billion at an $11 billion valuation to take on Nvidia
AI chip maker SambaNova raised $1 billion in a round led by General Atlantic, tripling its value months after Intel was rumoured to be circling it for a fraction of the price. Investors are funding real alternatives to Nvidia, which over time means more supply and lower costs for anyone running AI workloads.
9. Canada’s Bill C-36 takes on AI and privacy, but experts say gaps remain
Canada’s new privacy bill promises stronger protections, particularly for children, but experts argue it still leaves significant AI-specific risks unaddressed. Canadian organizations should read it as a floor rather than a ceiling — the compliance bar for AI systems here is still being set.
10. AI data labeller Mercor is in talks to raise $500 million at a $20 billion valuation
Mercor, which supplies the human expertise used to train and evaluate AI models, is reportedly raising at a $20 billion valuation — a number that would make its three 23-year-old founders billionaires several times over. The valuation is a signal worth reading: the scarce input in AI right now is not compute, it is qualified human judgment.
