AI Business Week: Nvidia, Anthropic and OpenAI Drive Chip, Deal and Workforce Moves

The week's AI business coverage centered on compute capacity and memory supply, with Nvidia, Anthropic, OpenAI and TSMC at the core of most stories. Chip shortages pushed companies toward new hardware, cash acquisitions and capacity contracts, while Nvidia's leadership change added a governance question to the list.

Chips and memory supply

Nvidia argued that memory pressure is rising as HBM supply tightens, and its full-stack codesign plan relies on NVFP4, Dynamo and KV cache tiering to reduce memory needs. (Nvidia's memory pressure pitch) TSMC has reportedly drafted 3D DRAM plans at Nvidia's request, yet TSMC says publicly that it stays focused on foundry, and no production or investment decision exists yet. (TSMC's 3D DRAM report)

Google added TPU 8tx and TPU 8i Max to its lineup, with claimed gains of 2.7x in training and 80% in inference price-performance over the prior generation. (Google's expanded TPU lineup) NVIDIA, SK Telecom and Rebellions plan R200 NPU systems on NVLink Fusion, with the SK Telecom deployment expected to be ready in 2028. (the Korean NPU rack plan)

TSMC's board approved up to $1 billion in investments in Anthropic and OpenAI, as AI chip demand outstrips its manufacturing capacity. (TSMC's investment in both labs)

AI business deals and acquisitions

Five9 agreed to buy Level AI for $325M in cash, about 47% of its cash pile, and expects 2027 EBITDA margin dilution. (Five9's Level AI purchase) Dropbox agreed to buy Tigris Data for about $350M in cash, a deal tied to its DBX Agent Cloud and AI storage offering. (Dropbox's Tigris Data acquisition)

Adobe and Google expanded their partnership, bringing Acrobat into Drive and Gmail and placing Google AI models inside Photoshop, Firefly and Premiere. (the expanded Adobe and Google partnership) AST SpaceMobile secured a Space Force contract worth about $100 million in space-based sensing revenue, with AST supporting prime contractor STR. (AST SpaceMobile's Space Force award)

Models, security and leadership

Anthropic's amended S-1 says the company may soon meet its recursive self-improvement threshold, while stating that no current model has crossed it. (Anthropic's S-1 disclosure) Anthropic's Lambda compute deal can now be terminated by either party after Lambda missed a Texas facility lease deadline, although Anthropic says it does not intend to terminate. (the Lambda compute agreement)

OpenAI released GPT-6 Cyber to selected Daybreak Red security teams with no public pricing, citing 81 Firefox flaws and 948 Linux kernel exploits. (OpenAI's restricted GPT-6 Cyber release) Jensen Huang resigned as Nvidia CEO, with CFO Colette Kress serving as interim CEO while Nvidia's AI deals and about $530B in commitments face scrutiny. (the Nvidia leadership change)

ServiceNow started another global layoff round across the US, UK and India, with $167M in nine-month severance costs and a flat year-end headcount goal. (ServiceNow's layoff round)

The next thing to watch is whether Nvidia's leadership transition changes how the company handles its roughly $530B in AI commitments.

Related Articles

• TSMC Investment in Anthropic and OpenAI Reaches Up to $1 Billion
• AI Business Roundup: Compute Buildout Meets Pacing Jitters as Agent Governance Hardens
• The Daily Brief: Chipmakers and Cloud Firms Line Up $26B in Anthropic Shares Ahead of IPO

✔Human Verified

Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.