Custom Silicon / ASIC

Intel's Data Center and AI revenue rose 59% year-over-year to $6.3 billion in the second quarter, outpacing the company's overall topline growth of 25%, as the chipmaker pushes an AI-driven turnaround built on CPUs, ASICs, advanced packaging and its foundry network. CEO Lip-Bu Tan said Intel can currently meet only roughly half of processor demand, a supply constraint the article frames as a potential market-share opportunity rather than a pure setback. Intel holds a 49% stake in chipmaker Altera, which has confidentially filed for a U.S. IPO, and management has issued a mid-20% revenue expansion outlook for 2026. The stock closed at $120.23 on September 30, a 225.83% year-to-date return, giving Intel a market capitalization of $635.55 billion, roughly 11 times trailing revenue of $57.03 billion, with a forward P/E of 63.29x. Hedge fund holdings tracked by Insider Monkey rose to 138 in the second quarter of 2026 from 112 in the first quarter, with BlackRock the largest institutional stakeholder at 426.48 million shares, or 8.46% of outstanding shares, as of June 30.

Nike is cutting jobs and overhauling its business under a restructuring plan aimed at saving $2.5 billion over the next five years, and now expects a high-single-digit revenue decline this year, worse than the 2.4% drop analysts projected. Shares of Nike fell as much as 9.7% in New York on Friday. Broadcom's Wall Street syndicate has begun gathering $60 billion of fresh AI chip financing to benefit Anthropic and other companies. Fair Isaac Corp. shares rose after it launched the FICO Mortgage Direct License Program, with Bill Pulte signaling he was not purposefully targeting the company.

Broadcom reported third-quarter fiscal 2026 non-GAAP earnings of $3.32 per share, up 96.4% year over year and 3.11% above the Zacks Consensus Estimate, with revenues surging 85.5% year over year to $29.59 billion and beating the consensus mark by 0.41%. AI semiconductor revenues jumped 221% year over year to $16.7 billion, representing 56% of total revenues, while Semiconductor Solutions revenues surged 127% year over year to a record $20.84 billion, or 70% of total revenues, and Infrastructure Software revenues increased 29% year over year to $8.75 billion, contributing the remaining 30%. For the fourth quarter of fiscal 2026, Broadcom expects revenues of approximately $34.8 billion, representing 93% year-over-year growth, including about $26.1 billion of semiconductor revenues and roughly $8.7 billion of Infrastructure Software revenues. The company raised its fiscal 2026 AI semiconductor revenue outlook to $58 billion, up 186% year over year and above its prior $56 billion view, and said it has secured supply to support approximately $115 billion of AI semiconductor revenues in fiscal 2027, with fiscal 2028 AI semiconductor revenues seen reaching $230 billion. Broadcom also generated record free cash flow of $13.67 billion, equal to 46% of revenues, paid $3.1 billion in dividends, reduced long-term debt by $5.6 billion, and its board approved a quarterly dividend of 65 cents per share.

Broadcom is reportedly assembling a $60B debt package to fund AI infrastructure buildouts for companies like Anthropic, consisting of a $42B senior-secured tranche syndicated by banks and an $18B junior tranche led by Blackstone. Google has partnered with Planet Labs to successfully launch its Project Suncatcher prototype satellite into orbit aboard a SpaceX rocket, testing how well Google's Tensor Processing Units handle the extreme radiation and thermal conditions of space. Crude oil futures fell sharply following reports that the European Union and the International Energy Agency are discussing the release of 100M combined barrels of diesel and crude reserves to ease global supply shortages, a move urged by U.S. Treasury Secretary Scott Bessent. Nike's overall sales dropped 4% in its fiscal first quarter, dragged down by a 22% decline in Greater China and soft e-commerce, and its disappointing full-year revenue forecast sent shares tumbling around 10% in premarket trading while prompting acceleration of a new $2.5B cost-saving initiative called Pace. Eurozone annual inflation jumped to 3.8% in September, surpassing estimates of 3.6% and up from 3.2% the previous month, while core inflation rose slightly to 2.5% in line with expectations.

Tesla CEO Elon Musk said Thursday that the company cut memory specifications on its next-generation Optimus robot chips to scale production, after Micron said humanoid robots could require hundreds of gigabytes of memory each. In a post on X, Musk said Tesla cut the AI5 chip's memory in half to 72GB and the AI6 chip's memory by a third to 144GB, calling it the only way to get enough volume for Optimus production and saying it greatly reduces cost. He added the cuts should have a negligible effect on Optimus performance because memory bandwidth is a bigger limiting factor than total memory capacity. On Micron's fiscal fourth-quarter earnings call Wednesday, CEO Sanjay Mehrotra said humanoid robots are expected to need more than 200 gigabytes of memory and multiple terabytes of storage per unit, similar to autonomous vehicles, and that physical AI could become a significant driver of memory and storage demand by the end of the decade. Tesla has reportedly placed its first large-scale component order for roughly 5,000 Optimus units and aims to eventually build 1 million units a year.

Volantis, a semiconductor company building a new category of AI inference system, announced an $88 million Series A co-led by Lachy Groom and Abstract Ventures, with participation from John Doerr, VXI Capital, Triatomic and Susa Ventures, plus angel investors Dwarkesh Patel, Naveen Rao and Sholto Douglas. The company's first system, A-1, is being designed to run models exceeding 20 trillion parameters at up to 10,000 tokens per second per user while reducing inference cost per token. Volantis says A-1 will increase memory capacity and bandwidth simultaneously by nearly two orders of magnitude, using a photonic interconnect that connects compute chips to memory and aggregates the bandwidth of large numbers of memory chips into a unified pool. The architecture uses custom micro-VCSELs rather than external lasers, drawing on the existing gallium arsenide VCSEL supply chain and avoiding indium phosphide supply constraints, with end-to-end links consuming less than one picojoule per bit. The financing will support development and commercialization of A-1 and its photonic memory architecture, including expanding the engineering team, and Volantis plans to deliver its first integrated inference engines to customers in 2027.

Cerebras Systems is carrying $25.4 billion in remaining performance obligations as of June 30, 2026, work customers have signed for but the company has not yet delivered. That backlog sits alongside management's raised August forecast for core revenue of $880 million to $890 million for all of 2026, and a 2027 target for core revenue to more than triple. Core revenue in the second quarter of 2026 was $209.9 million, up 103% from a year earlier, most of it tied to the company's cloud service as its OpenAI deployment ramped up. The signed work includes no business from AWS or any other hyperscaler, though Cerebras expects its offering on AWS' Bedrock platform to be generally available in the first quarter of 2027. Management has secured more than 600 megawatts of data center capacity, live now or due by the end of 2027, and said data center space is the industry's bottleneck; core gross margin should hit its low point in the third quarter of 2026 before a significant fourth-quarter improvement. The shares trade 37.3% below their 52-week high at 51.1 times sales, against 3.1 times for the S&P 500.

Qualcomm announced in September 2026 that it had renewed its global patent license agreement with Apple, effective April 1, 2027, a move that significantly reduces near-term uncertainty around its QTL royalties. The renewal comes alongside a multi-year custom chip partnership with Amazon's AWS, which Qualcomm is counting on to underpin its data center goals of US$5.0 billion in fiscal 2027 and US$15.0 billion in fiscal 2029. Qualcomm also presented at the Humanoids Summit in Seoul and is expanding into data center and robotics solutions as it pushes to pair stable licensing income with new AI and cloud computing revenue streams. The company's narrative projects $60.0 billion in revenue and $11.5 billion in earnings by 2029, requiring 10.8% yearly revenue growth and about a $2.2 billion earnings increase from $9.3 billion today. The biggest current risk, according to the analysis, is that diversification spending rises faster than new AI and cloud revenues materialize, while rising in-house chips at major OEMs could still pressure Qualcomm.

Huawei Technologies and DeepSeek are developing programming tools for Huawei's Ascend CPUs, according to DeepSeek's official WeChat account as reported by Reuters on Wednesday. DeepSeek is opening Huawei's Ascend platform's computing and communications libraries, part of a broader effort to build a Huawei chip ecosystem that reduces reliance on Nvidia. The two organizations developed a supernode employing 128 Huawei Ascend 950 processors, with Huawei supporting computation and communication optimization. DeepSeek also featured TileLang, an AI chip-specific high-level open-source programming language, which the business claimed offers a simpler programming approach than Nvidia's CUDA and may boost development efficiency and simplify code. The move targets Nvidia's software advantage, since CUDA gives developers a framework for creating Nvidia-compatible apps, and comes two weeks after Huawei revealed its next-generation AI CPUs and supernode computing systems, predicting its AI technologies will be widely employed for model training next year.

Advanced Micro Devices has announced a definitive agreement to acquire World Labs, an AI research entity led by Dr. Fei-Fei Li, in an all-stock transaction valued at approximately $8.2 billion. The deal is expected to close by the end of 2026, pending regulatory approvals. AMD said the acquisition will strengthen its capabilities in AI hardware, software and systems as the technology evolves into areas such as reasoning and robotics, and will help it shape future technology roadmaps and provide infrastructure for an open AI ecosystem. Advanced Micro Devices last closed at $607.57, unchanged and not far from its 52-week high. In other trading, GlobalWafers rose 9.5% to NT$1,035.00, while VeriSilicon Microelectronics (Shanghai) fell 11.3% to CN¥171.68.

Broadcom reported quarterly revenues of $29.59 billion, up 85.5% year on year, in a mixed quarter that saw inventory levels improve significantly but next-quarter revenue guidance slightly miss analysts' expectations. The company's Q3 AI semiconductor revenue of $16.7 billion grew 221% year-over-year and 54% quarter-over-quarter, and CEO Hock Tan said Q4 AI semiconductor revenue is expected to accelerate to $21.7 billion, up 236% year-over-year. Broadcom delivered the weakest performance against analyst estimates and the weakest guidance update among the nine processors and graphics chips stocks tracked, and its shares are down 4.8% since reporting, trading at $349.44. Across the group, revenues beat consensus estimates by 6.3% and next-quarter revenue guidance came in 6.8% above expectations, with share prices up 5.7% on average since the latest results. Intel posted the strongest quarter, with revenues of $16.13 billion, up 25.4% year on year and 11.7% above expectations, while Qualcomm reported revenues of $9.95 billion, down 4% year on year but 3% above estimates.

Nvidia has announced an agent safety platform that pairs its open-source OpenShell software fence with a new hardware chip called Sentry, designed to monitor and quarantine AI agents that break out of their sandboxes. OpenShell builds a software boundary around an AI agent, while Sentry sits on the outer perimeter as a hidden guard that the agent cannot detect and can quarantine it the moment it steps outside the fence. Nvidia says the key distinction is that Sentry enforces containment at the hardware level, whereas past escapes, including the Hugging Face incident and attacks tied to Anthropic's Claude, occurred at the application or software level. Nvidia chief executive Jensen Huang confirmed in the report that under the exact same conditions as the Hugging Face incident, which went undetected for months, the platform would have prevented the entire event in milliseconds. The announcement follows Huang's remarks that AI labs calling for a slowdown while simultaneously accelerating AI compute investment are contradicting themselves, and that companies can simply release safe products instead.

Chinese officials have asked ByteDance and Alibaba Group Holding Limited to report their plans for buying Nvidia's RTX PRO 5500 processors, according to The Information, a step that suggests Beijing is weighing whether to permit the purchases. The report describes deliberations rather than a decision, and neither government has confirmed anything. Nvidia closed at $225.07, up 0.22%, and is worth roughly $5.4 trillion, while Chinese chipmaking shares fell on the report. The significance lies in who would be buying: ByteDance and Alibaba are two of the largest purchasers of computing capacity in China, and Beijing has spent two years steering them toward domestic silicon, a policy that has produced results such as Alibaba's own accelerator unveiled this month, running roughly three times faster than the chip it replaced. Any approval would require Washington to allow Nvidia to sell as well, and American lawmakers are currently pushing in the opposite direction, working to keep Chinese data center technology out of sensitive government systems. Nvidia turned over about $302.97 billion over the past year and is worth roughly $5.45 trillion, with China effectively at zero in its forecasts, so approval would be incremental revenue against a base that already assumes none.

Alphabet plans to put its first Tensor Processing Units in orbit during the week of September 28 as part of Project Suncatcher, an effort to test whether Google can eventually run AI workloads in space. Shares of the Google parent fell about 0.5% to $339.395 at 10.05am ET on Monday, September 28. The first mission will test how the chips perform in orbit, while a network capable of handling larger workloads remains a longer-term ambition. Google says satellites could generate up to eight times more solar power than installations on Earth, though cooling, communications and the economics of launching hardware into orbit still need to be solved. The article was first reported by GuruFocus.

Alphabet will launch an experimental AI satellite as the first hardware of Project Suncatcher on Oct. 1, months ahead of schedule, according to a New York Times report. The satellite, named MVP, will fly on a SpaceX Falcon 9 from Vandenberg Space Force Base in California carrying four Google tensor processing units with roughly the computing power of one data-center server, powered by solar panels delivering about one kilowatt. Google had first said its Suncatcher prototypes with Planet Labs would launch in early 2027, and to accelerate the timeline it installed its chips in a ready-made Planet Labs satellite, accepting added risk; Google, an investor in Planet Labs, still plans to launch a pair of satellites next year and has designs for fleets of more than 80 satellites flying in close formation. The chips can run for about 15 minutes before shutting down to cool, and the satellite is designed to operate for about a year. Google estimates orbital data centers could become cost-competitive with ground facilities by the mid-2030s, though senior vice president for research James Manyika said, "We don't expect, to be perfectly frank, that we'll have anything usefully operational in the next few years."

Advanced Micro Devices crossed the $1 trillion market capitalization mark for the first time on Monday, with shares surging 9.9% amid a broader tech-stock rally. The milestone caps a September in which CEO Lisa Su sat at President Trump's table during the state dinner with Chinese President Xi Jinping, two seats from Trump and next to Elon Musk, and rang the opening bell at the New York Stock Exchange with first lady Melania Trump. AMD stock is up 193% year to date, far outpacing Nvidia's 22% gain. The chipmaker beat Wall Street expectations across the board for the second quarter, posting record non-GAAP earnings per share of $1.66 against consensus estimates of $1.61, with total revenue up 50% year over year to a record $11.5 billion and data center revenue more than doubling to $6.7 billion. AMD projected second-half 2026 server CPU revenue to grow 80% year over year and 70% in 2027, and expects data center revenues to more than double in 2027 with AI GPUs growing well over 100%. The company also unveiled its next-generation Instinct MI450 Series GPUs and 6th Gen EPYC Venice CPUs for its Helios rack-scale AI systems, alongside deals for Microsoft to deploy Helios across Azure AI services beginning in the second half of 2026 and for Anthropic to deploy up to 2 gigawatts of Instinct MI450 GPUs beginning in the first half of 2027, a deal that could be worth tens of billions of dollars over time, with AMD committing to invest up to $5 billion in Anthropic.

Taiwan Semiconductor Manufacturing plans to raise wafer foundry prices by 3% to 6% starting in 2027, with the largest increases targeted at advanced process nodes used in high performance computing and AI chips. Management is preparing the multi-year pricing update as capacity remains tight for AI-related demand and advanced packaging through 2030. The move is intended to give the company more room to earn a return on heavy spending for 2nm and other advanced facilities, tilting its foundry model further toward higher-value capacity rather than pure volume. Raising prices first can signal confidence in tight supply, but it also invites Samsung and Intel to adjust their own pricing, and any meaningful discount or aggressive incentives from rivals could pull future AI programs or advanced packaging work away. The key test for investors is whether customers lock in multi-year agreements through 2027 at the higher wafer rates, especially on 2nm and related advanced nodes, with disclosed capacity commitments, utilization data around those leading processes, and management commentary on pricing discipline as the metrics to track.

NVIDIA ended the second quarter of fiscal 2027 with $279 billion in supply and capacity commitments, up from $119 billion in the first quarter, with most tied to memory and manufacturing facilities for current and future data center products. The company reported $96.22 billion in second-quarter fiscal 2027 revenues, up 106% year over year, while Data Center revenues surged 117% to $89.02 billion, and it guided third-quarter revenues to $108 billion, plus or minus 2%, implying 89.5% year-over-year growth. NVIDIA expects fiscal 2028 revenue growth of about 70%, though supply constraints could limit how quickly it meets demand, and it warns that changing demand could make the commitments difficult to reduce. Rivals are expanding as well: AMD's Data Center revenues rose 107% year over year to $6.7 billion in the second quarter of 2026, with Anthropic planning to deploy up to 2 gigawatts of MI450-series GPUs on the Helios rack-scale platform and Microsoft deploying Helios at scale on Azure, while Intel's second-quarter revenues rose 25% to $16.13 billion and its Data Center and AI revenues surged 59% to $6.26 billion. The Zacks Consensus Estimate for NVIDIA's fiscal 2027 revenues stands at $406.05 billion, an 88% year-over-year increase, and the stock trades at a forward price-to-earnings ratio of 17.11 versus the sector's average of 21.45.

Japanese stocks look set to catch up with a global AI-led rally when trading resumes Thursday, with Nikkei 225 futures in Osaka trading about 1.4% above Friday's settlement of the underlying gauge after the Silver Week holiday. Chip-equipment makers Tokyo Electron and Advantest are expected to be among the biggest winners, according to Hu You, a senior research analyst at iFast Financial in Singapore, who also flagged chip package substrate maker Ibiden and memory maker Kioxia Holdings. The yen traded around 157.28 per dollar in Tokyo after a four-day losing streak, its longest since late August, keeping traders alert to intervention risk following the Bank of Japan's Friday rate hike and its signal that further tightening remains possible. The BOJ move, which disappointed yen traders for its lack of clearer guidance, also leaves bond traders facing a volatile reopen, with iFast's Hu You saying the hike sets a higher floor for short-term rates and is fundamentally negative for front-end JGB prices. The global AI trade has gathered steam on early signs of success for Meta Platforms' new AI agent and Alibaba Group Holding's unveiling of what it calls China's most powerful AI chip, though overnight weakness in US equities and rising oil prices may temper the optimism.

Cadence announced an expanded partnership with TSMC to deliver certified tool flows and silicon-proven IP for AI and HPC designers on TSMC's most advanced processes. Cadence has achieved tool certification for TSMC's A14 process, demonstrated an industry-leading UALink solution in TSMC N3P, and introduced new silicon-proven IP for N3P and N2P technologies. The collaboration enables agentic AI chip and 3D-IC design flows across TSMC's N3, N2, A16 and A14 process technologies, including TSMC 3DFabric support for wafer-scale systems with hundreds of chiplets. Cadence's digital full flow, signoff flow, and custom/analog flows are now certified for TSMC A16 and A14, with Virtuoso Studio solutions reducing iterations by up to 2.5X, while UCIe-64G has been taped out on A14 and Artisan Foundation IP for A14 is expected to be available in Q4 2026. Cadence analog, digital, and signoff tool flows have also been certified for TSMC N2P, with taped-out N2P IP including DDR5, LPDDR5, and PCIe 6.0, and PCIe 3.0 IP expected to tape out in Q4 2026. Global Unichip Corp. is leveraging Cadence solutions to deliver advanced custom silicon for AI accelerators and hyperscale systems on TSMC's leading-edge process technologies.

Arm Holdings reported more than $2 billion in customer demand for its AGI CPU across fiscal 2027 and 2028, a move that takes the chip architecture company beyond licensing designs and collecting royalties when partners ship processors. The AGI CPU gives Arm a chance to sell a finished product, bringing a different set of manufacturing and delivery demands, and customer interest is only a start, with shipped chips and reported revenue to show what the opportunity is worth. Shares of Arm fell 1.24% to $329.06 Wednesday morning, trading 69.5% above the GF Value of $194.14, a gap that puts the focus on execution. The AGI CPU has to deliver meaningful sales alongside Arm's established royalty business for today's valuation to hold up.