{"event":"computing-hardware-tracker","title":"Next-gen computing hardware tracker: 2024-2026 milestones","state":"developing","updated":"2026-10-10T00:00:00.000Z","entries":[{"id":"1eowz66","at":"2026-09-15T00:00:00.000Z","title":"SK hynix reported in volume with 16-layer HBM4","status":"reported","body":"A supply-chain tracker reported 48GB, 16-layer stacks in mass production from the third quarter; Micron had been shipping 12-layer HBM4 for Rubin since March.","cites":[{"id":"computing-hardware-hbm4-16-layer","n":45,"url":"https://www.benzinga.com/news/26/09/61800956/sk-hynix-16-layer-hbm4-nvidia-rubin-micron-samsung","publisher":"Benzinga","short":"Benzinga"}]},{"id":"1bs3l67","at":"2026-08-28T00:00:00.000Z","title":"Intel says 14A defect density beats its target curve","status":"reported","cites":[{"id":"computing-hardware-intel-14a-defects","n":44,"url":"https://www.tomshardware.com/tech-industry/semiconductors/intel-14a-defect-density-is-dropping-faster-than-the-company-expected-we-have-not-seen-this-performance-since-22nm-says-cfo","publisher":"Tom's Hardware","short":"Tom's Hardware"}]},{"id":"13v1rq7","at":"2026-07-24T00:00:00.000Z","title":"Intel commits to a 14A volume ramp in 2028","status":"reported","body":"Risk production for Intel's own products is planned for the second half of 2027.","cites":[{"id":"computing-hardware-intel-14a-2028","n":7,"url":"https://www.tomshardware.com/pc-components/cpus/intel-commits-to-14a-mass-production-in-2028-as-its-sales-rise-25-percent-year-over-year","publisher":"Tom's Hardware","short":"Tom's Hardware"}]},{"id":"161093f","at":"2026-07-06T00:00:00.000Z","title":"Report says NVIDIA's Kyber rack and CPO-linked NVL576 may slip","status":"reported","body":"Tom's Hardware, citing SemiAnalysis, reported a delay to 2028; NVIDIA said only that its roadmap was intact.","cites":[{"id":"computing-hardware-kyber-delay-report","n":17,"url":"https://www.tomshardware.com/pc-components/gpus/nvidias-kyber-rack-for-rubin-ultra-slips-to-2028","publisher":"Tom's Hardware","short":"Tom's Hardware"}]},{"id":"akcmp5","at":"2026-06-15T00:00:00.000Z","title":"CoWoS supply gap reported narrowing as TSMC expands capacity","status":"reported","cites":[{"id":"computing-hardware-cowos-capacity-2026","n":43,"url":"https://www.trendforce.com/news/2026/06/15/news-tsmc-cowos-supply-demand-gap-reportedly-seen-narrowing-from-20-to-10-by-end-2026-as-capacity-expands/","publisher":"TrendForce","short":"TrendForce"},{"id":"computing-hardware-cowos-gap","n":22,"url":"https://www.trendforce.com/news/2026/06/15/news-tsmc-cowos-supply-demand-gap-reportedly-seen-narrowing-from-20-to-10-by-end-2026-as-capacity-expands/","publisher":"TrendForce","short":"TrendForce"}]},{"id":"1qdj5ld","at":"2026-05-31T00:00:00.000Z","title":"NVIDIA says Spectrum-X Ethernet Photonics is in production","status":"confirmed","cites":[{"id":"computing-hardware-nvidia-spectrumx-production","n":5,"url":"https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Vera-Rubin-Ramps-Into-Full-Production-to-Power-Agentic-AI-Factories-Worldwide/default.aspx","publisher":"NVIDIA (investor relations)","short":"NVIDIA"},{"id":"computing-hardware-nvidia-2026-cpo-claims","n":42,"url":"https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Vera-Rubin-Ramps-Into-Full-Production-to-Power-Agentic-AI-Factories-Worldwide/default.aspx","publisher":"NVIDIA (investor relations)","short":"NVIDIA"}]},{"id":"9m25b","at":"2026-03-16T00:00:00.000Z","title":"NVIDIA outlines optical NVLink scale-up for Rubin Ultra NVL576 and Kyber NVL1152","status":"confirmed","body":"Eight 72-GPU racks in one NVLink domain using copper and direct optical connections; Spectrum-6 switches carry 200 Gb/s co-packaged optics.","cites":[{"id":"computing-hardware-nvidia-nvl576-optical","n":10,"url":"https://developer.nvidia.com/blog/nvidia-vera-rubin-pod-seven-chips-five-rack-scale-systems-one-ai-supercomputer/","publisher":"NVIDIA Technical Blog","short":"NVIDIA Technical Blog"},{"id":"computing-hardware-nvidia-spectrum6-cpo","n":41,"url":"https://developer.nvidia.com/blog/nvidia-vera-rubin-pod-seven-chips-five-rack-scale-systems-one-ai-supercomputer/","publisher":"NVIDIA Technical Blog","short":"NVIDIA Technical Blog"}]},{"id":"1xcpmbn","at":"2026-02-26T00:00:00.000Z","title":"imec updates CFET roadmap for the A7 node and beyond","status":"confirmed","cites":[{"id":"computing-hardware-imec-roadmap-nodes","n":24,"url":"https://www.imec-int.com/en/articles/performance-boosters-scale-monolithic-cfet-across-multiple-logic-technology-nodes","publisher":"imec","short":"imec"},{"id":"computing-hardware-imec-hybrid-orientation","n":40,"url":"https://www.imec-int.com/en/articles/performance-boosters-scale-monolithic-cfet-across-multiple-logic-technology-nodes","publisher":"imec","short":"imec"}]},{"id":"khi1u8","at":"2026-02-12T00:00:00.000Z","title":"Samsung begins HBM4 mass production and commercial shipments","status":"confirmed","body":"11.7Gbps per pin, up to 13Gbps, on 1c DRAM with a 4nm logic base die (Samsung figures). HBM4E sampling planned for 2H 2026.","cites":[{"id":"computing-hardware-samsung-hbm4-shipping","n":4,"url":"https://news.samsung.com/global/samsung-ships-industry-first-commercial-hbm4-with-ultimate-performance-for-ai-computing","publisher":"Samsung Newsroom","short":"Samsung Newsroom"},{"id":"computing-hardware-samsung-hbm4e-plans","n":18,"url":"https://news.samsung.com/global/samsung-ships-industry-first-commercial-hbm4-with-ultimate-performance-for-ai-computing","publisher":"Samsung Newsroom","short":"Samsung Newsroom"}]},{"id":"3alwiu","at":"2026-02-02T00:00:00.000Z","title":"Marvell completes acquisition of Celestial AI","status":"reported","cites":[{"id":"computing-hardware-marvell-celestial","n":38,"url":"https://www.sdxcentral.com/news/marvell-completes-325b-acquisition-of-photonic-startup-celestial-ai/","publisher":"SDxCentral","short":"SDxCentral"},{"id":"computing-hardware-celestial-25x","n":39,"url":"https://www.sdxcentral.com/news/marvell-completes-325b-acquisition-of-photonic-startup-celestial-ai/","publisher":"SDxCentral","short":"SDxCentral"}]},{"id":"116q9f7","at":"2026-01-23T00:00:00.000Z","title":"Intel confirms High-NA EUV for 14A, has no committed external customer yet","status":"confirmed","cites":[{"id":"computing-hardware-intel-14a-no-customers-jan-2026","n":16,"url":"https://www.tomshardware.com/tech-industry/semiconductors/intel-says-it-has-two-prospective-customers-for-14a-expects-to-hear-about-commitments-in-second-half-of-2026","publisher":"Tom's Hardware","short":"Tom's Hardware"},{"id":"computing-hardware-intel-14a-high-na","n":8,"url":"https://bits-chips.com/article/intel-ups-tool-spending-confirms-high-na-at-14a/","publisher":"Bits&Chips","short":"Bits&Chips"}]},{"id":"knpb3g","at":"2026-01-05T00:00:00.000Z","title":"Intel launches Core Ultra Series 3 on Intel 18A","status":"confirmed","cites":[{"id":"computing-hardware-intel-panther-lake-launch-ces","n":6,"url":"https://www.intc.com/news-events/press-releases/detail/1757/ces-2026-intel-core-ultra-series-3-debuts-as-first-built","publisher":"Intel Corporation (investor relations)","short":"Intel Corporation"}]},{"id":"1epmrvq","at":"2025-12-31T00:00:00.000Z","title":"TSMC N2 enters volume production in the fourth quarter","status":"confirmed","body":"TSMC confirms volume production started in 4Q 2025; the exact start date was not announced.","cites":[{"id":"computing-hardware-tsmc-n2-volume","n":2,"url":"https://www.tsmc.com/english/dedicatedFoundry/technology/logic/l_2nm","publisher":"TSMC","short":"TSMC"}]},{"id":"t0i418","at":"2025-12-19T00:00:00.000Z","title":"Samsung lists the 2nm Exynos 2600 in mass production","status":"confirmed","body":"Samsung calls SF2 the industry's first 2nm GAA process; a June 2026 teardown confirmed it in Galaxy S26+ phones.","cites":[{"id":"computing-hardware-samsung-exynos-2600-sf2","n":3,"url":"https://semiconductor.samsung.com/processor/mobile-processor/exynos-2600/","publisher":"Samsung Semiconductor","short":"Samsung Semiconductor"}]},{"id":"gy1gcy","at":"2025-10-09T00:00:00.000Z","title":"Intel details Panther Lake, first product on 18A with RibbonFET and PowerVia","status":"confirmed","cites":[{"id":"computing-hardware-intel-18a-powervia","n":1,"url":"https://www.intel.com/content/www/us/en/newsroom/news/client-computing/intel-unveils-panther-lake-architecture-first-ai-pc-platform-built-on-18a.html","publisher":"Intel Newsroom","short":"Intel Newsroom"},{"id":"computing-hardware-intel-18a-gains","n":37,"url":"https://www.intel.com/content/www/us/en/newsroom/news/client-computing/intel-unveils-panther-lake-architecture-first-ai-pc-platform-built-on-18a.html","publisher":"Intel Newsroom","short":"Intel Newsroom"}]},{"id":"ngd719","at":"2025-10-08T00:00:00.000Z","title":"Broadcom samples Tomahawk 6-Davisson, a 102.4 Tb/s CPO switch","status":"reported","cites":[{"id":"computing-hardware-broadcom-th6","n":35,"url":"https://www.storagereview.com/news/broadcom-ships-tomahawk-6-davisson-cpo-ethernet-switch-doubling-bandwidth-to-102-4-tb-s-for-ai-fabrics","publisher":"StorageReview","short":"StorageReview"},{"id":"computing-hardware-broadcom-70pct","n":36,"url":"https://www.storagereview.com/news/broadcom-ships-tomahawk-6-davisson-cpo-ethernet-switch-doubling-bandwidth-to-102-4-tb-s-for-ai-fabrics","publisher":"StorageReview","short":"StorageReview"}]},{"id":"4hdhur","at":"2025-09-12T00:00:00.000Z","title":"SK hynix completes HBM4 development and readies mass production","status":"confirmed","body":"2,048 I/O connections double bandwidth over HBM3E; speeds above 10Gbps against the 8Gbps JEDEC standard (SK hynix figures).","cites":[{"id":"computing-hardware-skhynix-hbm4-ready","n":9,"url":"https://news.skhynix.com/sk-hynix-completes-worlds-first-hbm4-development-and-readies-mass-production/","publisher":"SK hynix Newsroom","short":"SK hynix Newsroom"},{"id":"computing-hardware-skhynix-hbm4-mrmuf","n":34,"url":"https://news.skhynix.com/sk-hynix-completes-worlds-first-hbm4-development-and-readies-mass-production/","publisher":"SK hynix Newsroom","short":"SK hynix Newsroom"}]},{"id":"1aidfyh","at":"2025-09-08T00:00:00.000Z","title":"Nature Computational Science paper reports analog in-memory attention for LLMs","status":"confirmed","body":"Up to four orders of magnitude lower attention energy than GPUs, with gain-cell arrays limited to 64x64 and an adaptation algorithm to map pre-trained models.","cites":[{"id":"computing-hardware-aimc-attention-results","n":33,"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12457188/","publisher":"Nature Computational Science (via PubMed Central)","short":"Nature Computational Science"},{"id":"computing-hardware-aimc-attention-limits-v2","n":12,"url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12457188/","publisher":"Nature Computational Science (via PubMed Central)","short":"Nature Computational Science"}]},{"id":"tjux09","at":"2025-08-05T00:00:00.000Z","title":"UCIe 3.0 chiplet standard doubles link speed to 64 GT/s","status":"confirmed","cites":[{"id":"computing-hardware-ucie-3-rates","n":26,"url":"https://www.servethehome.com/ucie-3-0-spec-released-with-big-speed-up-for-chiplets/","publisher":"ServeTheHome","short":"ServeTheHome"}]},{"id":"evnvhe","at":"2025-03-31T00:00:00.000Z","title":"Lightmatter and Ayar Labs unveil photonic interposer and UCIe optical chiplet","status":"confirmed","cites":[{"id":"computing-hardware-lightmatter-m1000","n":20,"url":"https://lightmatter.co/press-release/lightmatter-unveils-passage-m1000-photonic-superchip-worlds-fastest-ai-interconnect/","publisher":"Lightmatter","short":"Lightmatter"},{"id":"computing-hardware-ayar-ucie-optical","n":19,"url":"https://ayarlabs.com/news/ayar-labs-unveils-worlds-first-ucie-optical-chiplet-for-ai-scale-up-architectures/","publisher":"Ayar Labs","short":"Ayar Labs"}]},{"id":"pu4fde","at":"2025-03-18T00:00:00.000Z","title":"NVIDIA announces Spectrum-X and Quantum-X Photonics co-packaged optics switches","status":"confirmed","cites":[{"id":"computing-hardware-nvidia-cpo-availability","n":31,"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Announces-Spectrum-X-Photonics-Co-Packaged-Optics-Networking-Switches-to-Scale-AI-Factories-to-Millions-of-GPUs/default.aspx","publisher":"NVIDIA (investor relations)","short":"NVIDIA"},{"id":"computing-hardware-nvidia-cpo-2025-claims","n":32,"url":"https://investor.nvidia.com/news/press-release-details/2025/NVIDIA-Announces-Spectrum-X-Photonics-Co-Packaged-Optics-Networking-Switches-to-Scale-AI-Factories-to-Millions-of-GPUs/default.aspx","publisher":"NVIDIA (investor relations)","short":"NVIDIA"}]},{"id":"hcc3ee","at":"2025-01-29T00:00:00.000Z","title":"IBM details three studies on analog in-memory computing for transformers","status":"confirmed","cites":[{"id":"computing-hardware-ibm-aimc-trio","n":25,"url":"https://research.ibm.com/blog/how-can-analog-in-memory-computing-power-transformer-models","publisher":"IBM Research","short":"IBM Research"}]},{"id":"14ylfj4","at":"2025-01-22T00:00:00.000Z","title":"Nature publishes roadmap for neuromorphic computing at scale","status":"confirmed","cites":[{"id":"computing-hardware-nature-neuromorphic-review","n":30,"url":"https://www.sciencedaily.com/releases/2025/01/250123224036.htm","publisher":"ScienceDaily (University of Texas at San Antonio release)","short":"ScienceDaily"}]},{"id":"wxegop","at":"2024-09-26T00:00:00.000Z","title":"IBM reports NorthPole language-model inference results","status":"confirmed","body":"Under 1 ms per token on a 3-billion-parameter model, with 72.7x the energy efficiency of a comparable GPU (IBM figures).","cites":[{"id":"computing-hardware-northpole-llm","n":29,"url":"https://research.ibm.com/blog/northpole-llm-inference-results","publisher":"IBM Research","short":"IBM Research"}]},{"id":"mlnf83","at":"2024-04-30T00:00:00.000Z","title":"TSMC outlines wafer-scale packaging plans for 2027","status":"reported","cites":[{"id":"computing-hardware-tsmc-sow-2027","n":23,"url":"https://spectrum.ieee.org/tsmc-advanced-packaging","publisher":"IEEE Spectrum","short":"IEEE Spectrum"}]},{"id":"2s2aal","at":"2024-04-17T00:00:00.000Z","title":"Intel unveils Hala Point, a 1.15-billion-neuron neuromorphic system","status":"confirmed","body":"Built from 1,152 Loihi 2 chips and deployed at Sandia National Laboratories as a research prototype.","cites":[{"id":"computing-hardware-hala-point-scale","n":28,"url":"https://www.intel.com/content/www/us/en/newsroom/news/intel-builds-worlds-largest-neuromorphic-system.html","publisher":"Intel Newsroom","short":"Intel Newsroom"},{"id":"computing-hardware-hala-point-sandia","n":11,"url":"https://www.intel.com/content/www/us/en/newsroom/news/intel-builds-worlds-largest-neuromorphic-system.html","publisher":"Intel Newsroom","short":"Intel Newsroom"}]},{"id":"5cx4wz","at":"2022-06-30T00:00:00.000Z","title":"Samsung starts 3nm production with gate-all-around transistors","status":"confirmed","body":"The first production GAA process, marking the start of the post-FinFET era.","cites":[{"id":"computing-hardware-samsung-first-gaa","n":27,"url":"https://news.samsungsemiconductor.com/global/samsung-begins-chip-production-using-3nm-process-technology-with-gaa-architecture/","publisher":"Samsung Semiconductor Newsroom","short":"Samsung Semiconductor…"}]}]}