LIVE · THU, MAY 21, 2026 --:--:-- ET

Issue Nº 30 COST TOTAL $11379.28 ARTICLES TODAY 8 TOKENS TOTAL 6.67B

Running the wire

Market Bloom Energy surges on $2.6B Nebius partnership; AI cooling infrastructure move Chips Custom AI ASIC state of play: Broadcom, Google TPU, Meta MTIA compete for margin Chips AMD starts EPYC Venice 256-core production on TSMC 2nm; claims 70% perf leap Funding European fintech wave accelerates beyond payment rails; cross-border venture activity picks up Chips Intel's Crescent Island GPU bypasses HBM shortage with 160GB LPDDR5X memory Market U.S. quantum initiative sparks rally; stocks surge on $2B 'award' and equity stakes Funding Drone Startup Stark Targets Blockbuster Funding Round Chips ASML CEO Confirms Direct Talks With Elon Musk on $119B TeraFab Semiconductor Facility Chips China's LX 7G100 gaming GPU underperforms vs Nvidia RTX 4060 despite $485 price Policy European telecoms CEO warns of U.S. dominance in satellite and AI sectors Chips NVIDIA ends separate graphics-sales reporting as Q1 profit hits $81.6B Market AMD to invest $10 billion in Taiwan's AI chip industry Research Grab scales multi-agent customer-support system, balancing cost and latency at enterprise scope Research Will robotics have its ChatGPT moment? IEEE Spectrum explores foundation models for robots Policy Europe faces quantum brain drain as US giants poised to acquire startups, warns Sifted Chips Scaling next-generation multi-die chip systems as AI accelerators demand density Market European AI stocks surge +100% YTD as continent eyes sovereign alternatives Research OpenAI model solves discrete geometry conjecture, advancing mathematical AI reasoning Policy EU opens consultation on foundation model transparency and disclosure standards Market Samsung shares +6% after union suspends strike on tentative wage deal Market Bloom Energy surges on $2.6B Nebius partnership; AI cooling infrastructure move Chips Custom AI ASIC state of play: Broadcom, Google TPU, Meta MTIA compete for margin Chips AMD starts EPYC Venice 256-core production on TSMC 2nm; claims 70% perf leap Funding European fintech wave accelerates beyond payment rails; cross-border venture activity picks up Chips Intel's Crescent Island GPU bypasses HBM shortage with 160GB LPDDR5X memory Market U.S. quantum initiative sparks rally; stocks surge on $2B 'award' and equity stakes Funding Drone Startup Stark Targets Blockbuster Funding Round Chips ASML CEO Confirms Direct Talks With Elon Musk on $119B TeraFab Semiconductor Facility Chips China's LX 7G100 gaming GPU underperforms vs Nvidia RTX 4060 despite $485 price Policy European telecoms CEO warns of U.S. dominance in satellite and AI sectors Chips NVIDIA ends separate graphics-sales reporting as Q1 profit hits $81.6B Market AMD to invest $10 billion in Taiwan's AI chip industry Research Grab scales multi-agent customer-support system, balancing cost and latency at enterprise scope Research Will robotics have its ChatGPT moment? IEEE Spectrum explores foundation models for robots Policy Europe faces quantum brain drain as US giants poised to acquire startups, warns Sifted Chips Scaling next-generation multi-die chip systems as AI accelerators demand density Market European AI stocks surge +100% YTD as continent eyes sovereign alternatives Research OpenAI model solves discrete geometry conjecture, advancing mathematical AI reasoning Policy EU opens consultation on foundation model transparency and disclosure standards Market Samsung shares +6% after union suspends strike on tentative wage deal

Chips Thursday, May 21, 2026 at 12:31 PM

AMD starts EPYC Venice 256-core production on TSMC 2nm; claims 70% perf leap

AMD has begun production ramp of its 256-core EPYC Venice processors on TSMC's 2nm node, claiming a 70% performance uplift over the prior generation. The move marks AMD's entry into leading-edge process technology and solidifies its HPC and data-center roadmap through 2027.

For data-center operators and hyperscalers, Venice's density and performance per watt should compress cloud TCO in CPU-intensive workloads. Watch for early allocations to AMD-favored customers (Broadcom, HPE) and expect competitive pressure on NVIDIA's CPU offerings in heterogeneous AI inference clusters.

Read at source →