31 Aug 2026Compute & Infrastructure
DASC compresses hybrid-attention state checkpoints while cutting serving latency
DASC exploits the decay structure of recurrent linear-attention states to compress checkpoints used for prompt reuse. On Kimi-Linear's KDA state, the authors report 2.63 times compression, 42.6% lower mean time-to-first-token and 68.4% higher input throughput, with a similar but smaller trend on Qwen GDN. The result is a first-party serving benchmark on a limited set of hybrid architectures and an eight-GPU setup.
PRIMARY CONFIRMEDPREPRINTPrimary source ↗ Relevance7.8/10evidence 82/100
31 Aug 2026Compute & Infrastructure
EuroHPC awards Bull a €387.8M contract for the LUMI-AI supercomputer
Bull was selected by EuroHPC and the LUMI AI Factory consortium to deliver LUMI-AI in Finland under a €387.8 million procurement contract. The system is designed to expand AI capacity tenfold and is scheduled for deployment in the second half of 2027, making this a committed infrastructure award rather than operational capacity.
INDEPENDENTLY CONFIRMEDBULLPrimary source ↗ Relevance8.3/10evidence 97/100
31 Aug 2026Compute & Infrastructure
HUMAIN and DataVolt begin construction of a 100MW AI data-center phase at NEOM
HUMAIN and DataVolt expanded their partnership and began construction of 100MW within the first 360MW phase of DataVolt's planned AI-ready campus at Oxagon, NEOM. The first 100MW is expected in 2028, so the event marks execution progress rather than operational compute capacity.
INDEPENDENTLY CONFIRMEDNEOMPrimary source ↗ Relevance8.4/10evidence 96/100
31 Aug 2026Compute & Infrastructure
SLB agrees to acquire Kelvion for $4.1B to scale AI data-center cooling
SLB signed an agreement to acquire thermal-management specialist Kelvion for $3.4 billion in cash plus $0.7 billion of assumed debt. The transaction targets AI data-center cooling capacity but is expected to close in the first half of 2027 subject to conditions and regulatory approvals.
INDEPENDENTLY CONFIRMEDSLBPrimary source ↗ Relevance8.2/10evidence 96/100
31 Aug 2026Compute & Infrastructure
NVIDIA invests $3.5B in MediaTek and expands NVLink Fusion partnership
NVIDIA invested $3.5 billion in MediaTek convertible bonds as the companies expanded collaboration across cloud AI factories, local AI and automotive computing. MediaTek will adopt NVLink Fusion for custom XPUs; the agreement is an announced ecosystem expansion, not evidence of newly installed capacity.
INDEPENDENTLY CONFIRMEDNVIDIA NEWSROOMPrimary source ↗ Relevance8.5/10evidence 96/100
28 Aug 2026Compute & Infrastructure
AutoDRI automates complex chip design-rule integration with near-perfect benchmark correctness
Researchers from Fudan University and UC San Diego report a multi-agent system that translates natural-language semiconductor design rules into executable CP-SAT constraints. Across 41 standard-cell benchmarks and 11 complex rules, the full workflow achieved 33/33 correct integrations with Gemini 3 Pro and 32/33 with GPT-5.4, with generated layouts passing KLayout DRC and Cadence LVS in the reported virtual-technology evaluations.
PRIMARY CONFIRMEDPREPRINTPrimary source ↗ Relevance7.8/10evidence 69/100
27 Aug 2026Compute & Infrastructure
NVIDIA pauses some AI-cloud revenue-sharing deals amid control and antitrust concerns
The Wall Street Journal reports that NVIDIA paused some agreements in its AI Compute Partnership program, which paired credit support and capacity backstops with a share of AI-cloud revenue. NVIDIA says the broader compute-access business model remains in place and is evolving.
PRELIMINARYWALL STREET JOURNALPrimary source ↗ Relevance7.3/10evidence 78/100
27 Aug 2026Compute & Infrastructure
SK hynix breaks ground on $4B Indiana HBM packaging facility
SK hynix has begun construction of its advanced HBM packaging and AI-semiconductor R&D facility in Indiana. The company now targets HBM4E volume production in Q3 2029 and eventual annual capacity of hundreds of thousands of wafers, but the site is not operational and depends on Korean-made wafers.
INDEPENDENTLY CONFIRMEDFIRST-PARTY SOCIALPrimary source ↗ Relevance7.2/10evidence 92/100
26 Aug 2026Compute & Infrastructure
AWS and NVIDIA plan 2 million additional GPUs across AWS in 2027–2028
AWS and NVIDIA announced a major infrastructure expansion centered on 2 million additional Blackwell Ultra, Rubin and Rubin Ultra GPUs across AWS in 2027–2028, plus 100,000 GPUs for U.S. government AI factories. The commitment is a major compute-supply signal but remains a future deployment plan rather than installed capacity.
INDEPENDENTLY CONFIRMEDAWS / NVIDIAPrimary source ↗ Relevance8.6/10evidence 97/100
26 Aug 2026Compute & Infrastructure
Anthropic reportedly agrees to a $45B Nscale compute lease for planned West Virginia capacity
Reuters reports that Anthropic will spend about $45 billion over six years to lease roughly 460 MW of planned AI compute capacity from Nscale in West Virginia. The companies did not confirm the report, and the capacity is not yet operational.
PRELIMINARYPRESS / WIREPrimary source ↗ Relevance8.2/10evidence 72/100
26 Aug 2026Compute & Infrastructure
NVIDIA introduces NVHBM for NVLink Fusion custom AI accelerators
NVIDIA introduced NVHBM, a custom HBM implementation for NVLink Fusion partners that moves the memory controller into the HBM base die; NVIDIA claims up to 30% more bandwidth, 15% lower HBM power and more compute-die area versus standard HBM4e.
PRIMARY CONFIRMEDNVIDIA TECHNICAL BLOGPrimary source ↗ Relevance7.8/10evidence 86/100
25 Aug 2026Compute & Infrastructure
OpenAI Jalapeño posts first silicon benchmarks with 1.5–1.9x more AI work per watt
OpenAI reports first measured results for its Jalapeño custom inference system on SemiAnalysis InferenceX. Across GPT-OSS 120B, DeepSeek R1 670B and Kimi K2.5 1T, Jalapeño delivered 1.5–1.9x higher peak AI work per watt and 1.7–3.6x lower end-to-end latency than compared Blackwell systems, with deployment inside OpenAI planned by year-end.
PRIMARY CONFIRMEDOPENAIPrimary source ↗ Relevance8.6/10evidence 91/100
24 Aug 2026Compute & Infrastructure
NVIDIA Groq 3 LPX reaches production with 3,431 tok/s at 100K context
NVIDIA says Groq 3 LPX is entering production deployment; Artificial Analysis measured 3,431 output tokens/s on Gemma 4 31B at 100K input context, about 4x the fastest public endpoint cited by NVIDIA, while Nebius is the first AI cloud adopter.
INDEPENDENTLY CONFIRMEDNVIDIAPrimary source ↗ Relevance8.0/10evidence 94/100
24 Aug 2026Compute & Infrastructure
Intel details Diamond Rapids and Crescent Island architectures for agentic AI
At Hot Chips 2026 Intel disclosed concrete architecture specifications for next-generation agentic AI infrastructure: Diamond Rapids with up to 256 cores, 1.28 GB LLC, 16 memory channels at 12,800 MT/s and 128 PCIe Gen6/CXL 3.0 lanes; Crescent Island with 32 Xe cores, 256 XMX engines, up to 480 GB LPDDR5X and a 350 W air-cooled PCIe form factor.
PRIMARY CONFIRMEDINTEL NEWSROOMPrimary source ↗ Relevance7.3/10evidence 88/100
24 Aug 2026Compute & Infrastructure
NVIDIA Vera Rubin posts first on-silicon AgentX results at up to 30x throughput per MW
NVIDIA reports first on-silicon Vera Rubin NVL72 measurements on SemiAnalysis AgentX production-style coding-agent sessions, reaching up to 30x higher throughput per megawatt and up to 35x lower token cost than GB300 NVL72 on selected workloads. Results are NVIDIA-measured and pending SemiAnalysis review.
PRIMARY CONFIRMEDNVIDIAPrimary source ↗ Relevance8.4/10evidence 86/100
23 Aug 2026Compute & Infrastructure
Alibaba prices HK$80B equity placement dedicated to full-stack AI
Alibaba priced a HK$80 billion (about US$10.2B) placement of 710 million new shares and says 100% of net proceeds will fund full-stack AI capabilities, including AI infrastructure.
INDEPENDENTLY CONFIRMEDALIBABA GROUPPrimary source ↗ Relevance7.6/10evidence 96/100
22 Aug 2026Compute & Infrastructure
NVIDIA customers reportedly face >15% AI-server price hikes as memory costs surge
Bloomberg reports that some major NVIDIA customers have been told server prices will rise by more than 15% in many cases for systems shipping in early 2027, including Vera Rubin and Grace Blackwell configurations, primarily because of rising memory costs. Reuters reported the claim but said it could not independently verify it; NVIDIA had not commented.
ANNOUNCEDPRESS / WIREPrimary source ↗ Relevance7.0/10evidence 68/100
21 Aug 2026Compute & Infrastructure
Anthropic hires Google TPU veteran Amir Salek as it builds an in-house silicon effort
Anthropic confirmed that Amir Salek, who founded Google’s custom-chip program and led the first seven TPU generations, is joining its compute team as the company lays groundwork for its own semiconductor initiative. The move adds unusually deep accelerator-design expertise to Anthropic’s effort to reduce dependence on external compute suppliers and co-design hardware for Claude workloads.
INDEPENDENTLY CONFIRMEDBLOOMBERGPrimary source ↗ Relevance7.2/10evidence 90/100
21 Aug 2026Compute & Infrastructure
YMTC parent advances toward $4.9B Shanghai IPO to expand NAND capacity and R&D
Shanghai Stock Exchange accepted the STAR Market listing application of CCSH, parent of YMTC. The planned offering targets roughly RMB33B ($4.9B), with proceeds earmarked for production-line upgrades and next-generation NAND/storage R&D. AI datacenter demand and tight NAND supply are material context, but the event is a capital-and-capacity signal rather than new compute capability.
INDEPENDENTLY CONFIRMEDPRESS / WIREPrimary source ↗ Relevance7.1/10evidence 90/100
21 Aug 2026Compute & Infrastructure
Starcloud raises $250M to scale orbital AI compute with NVIDIA and Cisco backing
Starcloud announced a $250M Series A extension at a $2.3B post-money valuation, bringing total funding to $450M. NVIDIA and Cisco Investments joined the round. Starcloud says the capital will expand manufacturing, fund engineering work with NVIDIA on the Space-1 Vera Rubin Module, and secure future launch capacity for its orbital data-center roadmap.
INDEPENDENTLY CONFIRMEDPRESS / WIREPrimary source ↗ Relevance7.3/10evidence 92/100
21 Aug 2026Compute & Infrastructure
NVIDIA invests in Cloverleaf to accelerate powered-land development for AI data centers
NVIDIA made a minority investment in Cloverleaf Infrastructure and entered a strategic partnership to accelerate U.S. AI data-center development. Cloverleaf develops power-ready sites and says it has delivered multiple gigawatt-scale projects; the partnership will use NVIDIA DSX to optimize site, power, cooling and compute infrastructure decisions. Financial terms were not disclosed.
INDEPENDENTLY CONFIRMEDPRESS / WIREPrimary source ↗ Relevance7.3/10evidence 95/100
20 Aug 2026Compute & Infrastructure
Broadcom seeks $60B+ debt package for frontier-AI chip financing
Broadcom is reported to be in talks with lenders for more than $60B of debt, potentially reaching about $100B across tranches, to finance AI-chip deployments benefiting Anthropic and other frontier labs. The talks extend the $35B Broadcom/Apollo/Blackstone AI XPV platform announced in June.
ANNOUNCEDPRESS / WIREPrimary source ↗ Relevance7.2/10evidence 72/100
19 Aug 2026Compute & Infrastructure
Google expands custom-AI silicon partnership with Marvell
Marvell will help develop Google custom AI chips and granted Google warrants for up to 58.97 million shares worth about $12.2B if fully exercised. The arrangement could generate up to roughly $120B of Marvell revenue through fiscal 2033 if order targets are met, broadening Google TPU supply across processors, storage and networking.
INDEPENDENTLY CONFIRMEDPRESS / WIREPrimary source ↗ Relevance7.6/10evidence 95/100
18 Aug 2026Compute & Infrastructure
Marin launches a fully open 535B-A23B frontier-scale training run
The Marin community launched a 535.3B-total / 22.76B-active MoE training run with an 18T-token budget on 11 GB200 NVL72 racks, publishing the model configuration, scaling ladder, training telemetry and research process in public. The signal is transparency and open frontier-scale training infrastructure, not a demonstrated capability jump.
PRIMARY CONFIRMEDFORMAL / REPOSITORYPrimary source ↗ Relevance7.1/10evidence 95/100
18 Aug 2026Compute & Infrastructure
Cerebras launches CS-4 wafer-scale inference system for hyperscale AI
Cerebras introduced CS-4, a rack-scale system built from three WSE-3 Turbo processors. The company reports up to 30x faster inference than production GPU systems and up to 10x more throughput per watt than CS-3; first shipments are scheduled for Q3 2026. The strongest headline figures are partly based on internal benchmarking and projections, so they should not be treated as independently validated capability measurements.
PRIMARY CONFIRMEDCEREBRASPrimary source ↗ Relevance7.3/10evidence 86/100