At the COMPUTEX expo in Taipei today, NVIDIA CEO Jensen Huang delivered what may be the year's most consequential keynote in the semiconductor industry. But the RTX Spark launch was far from the only seismic event — the entire AI landscape shifted in several directions at once today.
RTX Spark: 1 Petaflop in a Laptop
The centerpiece of the GTC Taipei keynote was RTX Spark, a brand-new system-on-chip that delivers 1 petaflop of AI performance in slim Windows laptops and compact desktops. The chip pairs a Blackwell-based RTX GPU (6,144 CUDA cores) with a custom 20-core Grace CPU, developed in partnership with MediaTek and linked via NVLink.
"Everything we've learned over 33 years, distilled into one chip," Huang said, calling it the most significant reinvention of the PC in 40 years. NVIDIA announced a new lineup of AI-native Windows machines that run large language models and an agent runtime directly on the OS — autonomous assistants that can "see, understand, and act" across files, apps, and the web.
The announcement positions NVIDIA squarely against Intel and Apple in the personal-computing market, aiming to replicate the smartphone-era paradigm shift in PCs.
GitHub Copilot Goes Usage-Based — Developers Revolt
Effective today, GitHub Copilot officially transitioned to token-based billing. Pro subscribers ($10/month) get $10 in monthly AI credits; Business ($19/user) and Enterprise ($39/user) receive proportional allowances. Anything beyond that is metered at published API rates.
Developer reaction has been fierce. Community threads on GitHub accumulated hundreds of comments and nearly 900 downvotes. Screenshots circulating on Reddit and X show projected monthly costs jumping from tens of dollars to hundreds or even thousands for heavy agentic workflows. TechCrunch summed up the mood with its headline: "'What a joke': GitHub Copilot's new token-based billing spurs consternation among devs."
GitHub says the new model is "more predictable and aligned with actual inference costs," and promises reduced rate-limiting once metering is live. Business and Enterprise users received temporary promotional credits for June through August.
Anthropic's $65B Haul — Nearing a Trillion
Last Thursday, Anthropic announced the close of a $65 billion Series H round at a $965 billion post-money valuation — making it the most valuable AI startup on the planet, surpassing OpenAI's $852 billion March valuation. The round was led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia, with strategic participation from Micron, Samsung, and SK hynix.
Coming just months after a $30 billion Series G in February, the raise reflects an acceleration of capital deployment into frontier AI labs. Run-rate revenue crossed $47 billion, and the company is positioning for a potential IPO.
SoftBank's €75B French Bet
On the infrastructure front, SoftBank announced plans to invest up to €75 billion ($87 billion) in AI data centers across France, totaling 5 gigawatts of capacity. The initial commitment of €45 billion will deliver 3.1 GW by 2031 across sites in Hauts-de-France near Dunkirk. Partners include state utility EDF (nuclear-powered grid) and Schneider Electric. The plan was unveiled ahead of President Macron's Choose France summit.
Cognition/Devin Raises $1B
The AI coding agent market continues to heat up. Cognition, the startup behind Devin, raised $1 billion at a $26 billion valuation — up from $10.2 billion just eight months ago. ARR reached $492 million, and Devin now writes 89% of Cognition's own production code.
The Bottom Line
June 2026 begins with multiple inflection points colliding: NVIDIA is trying to redefine the PC as an AI platform while developers discover that the era of flat-rate unlimited AI tooling has ended. European infrastructure is accelerating on the back of SoftBank's nuclear-powered bet, and private capital continues to flow into frontier AI at historic pace. The question now is whether the industry can build the compute capacity fast enough to keep up.
