Intel and Qualcomm Circle Tenstorrent in AI Chip Consolidation Race
Acquisition talks signal chipmakers are now buying AI capabilities rather than building them as Nvidia's market dominance weakens.
Intel and Qualcomm are in early acquisition discussions with Tenstorrent, the AI chip startup led by legendary architect Jim Keller, marking a strategic shift as established chipmakers pursue external IP over in-house development.
The competition for Tenstorrent reflects three converging forces reshaping the semiconductor industry: accelerating M&A as companies race to acquire ready-made AI capabilities, weakening of Nvidia’s market grip as custom silicon gains traction, and geopolitical competition for domestic AI chip sovereignty. According to Bloomberg, both companies held conversations with the startup as of May 18, though no formal offers have been made.
Semiconductor industry M&A surged from $2.7 billion in 2023 to $45 billion in 2024, per MarketMinute. AI Chips are expected to account for over $500 billion of the projected $975 billion global chip market in 2026.
Why Tenstorrent Matters
Tenstorrent’s RISC-V architecture and software stack represent one of the few production-ready alternatives to Nvidia’s CUDA ecosystem. The company’s Blackhole accelerator, announced in August 2024, delivers 745 teraFLOPS of FP8 performance with Ethernet-based interconnect rather than proprietary networking, according to The Register. The company is building a 32-chip Blackhole Galaxy system for large-scale deployments.
The startup has raised over $1 billion from Bezos Expeditions, Samsung, LG Electronics, Hyundai Motor Group, and Fidelity. Beyond accelerators, Tenstorrent develops three chip categories: CPUs, inference accelerators, and custom ASICs following its acquisition of Blue Cheetah in July 2025.
“We are the only company I know of that uses its own IP product in its own chips. Customers actually want the IP we are using in our own products because we are betting our company on the IP.”
— Aniket Saha, Vice President of Product Strategy at Tenstorrent, per EE Times
Intel’s Gaudi Struggles Drive M&A Logic
Intel’s pursuit of Tenstorrent reflects disappointing traction for its Gaudi accelerator line. While the company claims Gaudi 3 delivers 70% better price-performance on Llama 3 80B inference versus Nvidia’s H100, according to Intel Newsroom, adoption remains limited. The company cancelled its Falcon Shores GPU program and is now targeting a late 2027 launch for Jaguar Shores with HBM4E memory.
Acquiring Tenstorrent would give Intel a differentiated RISC-V architecture, established software ecosystem, and customer relationships across hyperscalers and international governments. The startup’s collaboration with Japan’s Leading-edge Semiconductor Technology Center on chiplet-based edge AI accelerators demonstrates commercial traction beyond China-focused deployments.
Qualcomm’s Datacenter Ambitions
Qualcomm’s interest signals an aggressive push beyond mobile chips into datacenter AI. The company acquired Alphawave IP Group in January 2026 to secure high-speed connectivity intellectual property, per TechInsights. Its AI200 chips are expected to ship in 2026 with Saudi Arabia’s Humain as the first customer, followed by AI250 in 2027.
However, Qualcomm lacks proven datacenter market presence. Tenstorrent would provide immediate credibility through Keller’s reputation and the company’s existing hyperscaler relationships. The acquisition would position Qualcomm as a full-stack AI infrastructure provider rather than a mobile chipmaker attempting datacenter adjacency.
Geopolitical Chip Sovereignty
The acquisition race occurs against intensifying geopolitical competition for AI chip independence. The US is investing nearly $12 billion to replicate Taiwanese advanced chip production in Arizona, according to Foreign Policy. Export controls targeting China’s access to advanced Semiconductors have accelerated domestic development efforts across multiple nations.
Tenstorrent’s open RISC-V architecture offers strategic flexibility. Unlike Nvidia’s proprietary CUDA ecosystem or Intel’s x86 licensing constraints, RISC-V enables sovereign chip development without US technology dependencies. The company’s international partnerships—from Japan’s government-backed semiconductor centre to Korean conglomerates—demonstrate this appeal.
- Established chipmakers are shifting from internal development to external IP acquisition as speed-to-market trumps vertical integration
- Nvidia’s CUDA moat faces fragmentation as custom silicon and open architectures gain traction among hyperscalers and governments
- RISC-V’s openness positions it as the architecture of choice for AI chip sovereignty efforts outside US influence
- Semiconductor consolidation will accelerate as AI capabilities become the primary profit driver, forcing laggards to buy rather than build
What to Watch
Tenstorrent’s valuation will test how the market prices ready-made AI IP against development timelines. With over $1 billion already raised, the startup commands premium pricing. Intel’s manufacturing struggles and Qualcomm’s datacenter inexperience may push the final price above $5 billion.
Whether Nvidia responds with its own M&A to strengthen software ecosystem lock-in. The company’s 20-year CUDA investment spanning 4 million developers and 3,000 optimised applications remains its primary moat, but open alternatives are narrowing that gap.
How US regulators approach semiconductor consolidation under national security review. Tenstorrent’s international partnerships and open architecture may trigger CFIUS scrutiny regardless of acquirer nationality.
The timeline for Tenstorrent’s Blackhole Galaxy system reaching production. Customer deployments would significantly strengthen acquisition valuation and validate the company’s technical claims against Nvidia benchmarks.