Anthropic’s $6B AI shopping spree sparks a new chip-and-model power struggle—who wins next?

Anthropic, the creator of the Claude chatbot, is reportedly negotiating the acquisition of the AI startup Decart for about US$6 billion, according to O Globo. The deal talks signal that model developers are moving aggressively to expand capabilities, talent, and proprietary technology rather than relying solely on third-party tooling. In parallel, coverage highlights how Jensen Huang’s strategy—pioneering specialized computer chips for the AI boom—helped build Nvidia into the world’s most valuable company. Another report argues that “custom silicon” is accelerating as hyperscalers rush to design their own chips, potentially turning into a serious competitor to Nvidia’s offerings. Geopolitically, this cluster points to a strategic contest over who controls the AI stack: frontier model makers, chip designers, and the cloud/hyperscaler platforms that deploy compute at scale. If Anthropic’s acquisition of Decart is real and successful, it could strengthen a US-based AI ecosystem that competes with other frontier labs by tightening the loop between models and specialized capabilities. Meanwhile, the custom-silicon push suggests hyperscalers may reduce dependency on a single supplier, reshaping leverage in export-control-sensitive supply chains and procurement budgets. Nvidia benefits from the current demand wave, but the risk is that a portion of future inference and training spend migrates toward in-house silicon ecosystems, changing bargaining power across the market. Market and economic implications are immediate for semiconductor supply chains, AI infrastructure capex, and cloud procurement. If hyperscalers accelerate custom silicon, it can pressure Nvidia’s pricing power and alter revenue mix across data-center GPUs and networking components, even if overall AI demand remains strong. For investors, the narrative supports continued strength in AI compute beneficiaries while increasing volatility around names most exposed to “GPU-only” demand. Currency and macro effects are indirect but relevant: large US$6 billion-style deals reinforce dollar-denominated capital flows into US tech, while global supply constraints in advanced packaging and memory can keep input costs elevated. What to watch next is whether Anthropic confirms the Decart talks, whether regulators scrutinize the transaction for competition or national-security concerns, and how quickly integration milestones are announced. On the chip side, track evidence that hyperscalers are scaling custom silicon from prototypes to production workloads, including performance-per-watt and total cost-of-ownership comparisons versus Nvidia platforms. A key trigger would be any sign that major cloud providers shift meaningful training or inference capacity away from Nvidia GPUs, which would show up in procurement guidance and supply allocation patterns. Over the next 1–3 quarters, the market will likely reprice AI infrastructure risk based on deal certainty, custom-silicon deployment metrics, and any export-control or licensing changes affecting advanced compute components.

Geopolitical Implications

• 01

  Vertical integration across models and specialized hardware is becoming a strategic advantage.

• 02

  Custom silicon can reduce supplier leverage and complicate export-control-sensitive procurement.

• 03

  US-based AI consolidation may intensify pressure on chip incumbents and reshape bargaining power.

Key Signals

• —Deal confirmation or withdrawal on Anthropic–Decart
• —Production scaling announcements for hyperscaler custom silicon
• —Procurement guidance indicating GPU share shifts
• —Regulatory filings tied to competition or national security