The ecosystem, not just product launches: fabless designers, foundries, IDMs, equipment and materials, and automation inside the fab.
Chip Manufacturing
9 POINTS
Applied Materials, the largest US chip-equipment maker, reported record fiscal 2026 revenue of $34.45 billion and fourth-quarter revenue of $10.41 billion, up 53% year on year, with record quarterly free cash flow of $2.6 billion. CEO Gary Dickerson said AI is fuelling demand and that its semiconductor systems business is on track to grow more than 40% in calendar 2026. Applied also said Intel will join its new EPIC Center in Silicon Valley, billed as the world's largest collaborative chip-process R&D facility, as a founding partner on next-generation transistors, interconnects and packaging, with Kioxia and Besi joining as innovation partners.
Semiconductor Digest · November 13, 2026
Why it matters: Equipment makers see orders before the fabs are built, so a 53% jump in Applied's quarter is a hard signal that AI chip capacity spending is still accelerating.
Watchpoint: The 40%-plus growth figure is Applied's own forecast for its systems business in calendar 2026, not a reported result.
- Fiscal year ended October 25, 2026; reported in the November 13 coverage.
- Record fiscal-year revenue: $34.45 billion.
- Fourth-quarter revenue: $10.41 billion, up 53% year on year.
- Non-GAAP Q4 EPS: a record $4.16, up 92%.
- Record quarterly free cash flow of $2.6 billion.
- Largest quarterly buyback in more than eight years.
- Systems business on track for 40%-plus growth in calendar 2026 (company forecast).
- Intel joins the EPIC Center as a founding partner.
- Kioxia and Besi join as innovation partners.
9-point recap written from the linked reporting. The article itself is the source.
Chip Manufacturing
9 POINTS
Alibaba confirmed on November 10 that T-Head, its in-house chip designer, filed confidentially in September for a Hong Kong listing, with Alibaba keeping control. International Financing Review reported the offering could raise at least $2 billion, and valuation estimates reported for JPMorgan and Nomura range from $25 billion to $85 billion. Founded in 2018, T-Head makes Zhenwu processors for AI training and inference, Yitian server CPUs, and licenses Xuantie RISC-V designs. Alibaba says more than 60% of its computing capacity serves customers outside the group, often via rented Alibaba Cloud capacity, which makes T-Head's own revenue hard to separate.
TechNode · November 11, 2026
Why it matters: A listed T-Head would be China's most direct public bet on home-grown AI accelerators as US export controls keep Nvidia's best chips out.
Watchpoint: Investors still do not have a clean figure for how much revenue belongs to T-Head itself rather than to Alibaba Cloud.
- Alibaba confirmed the filing on November 10.
- Confidential Hong Kong filing made in September.
- Alibaba will retain control of T-Head.
- IFR: the IPO could raise at least $2 billion.
- Reported valuations: JPMorgan $25-62 billion, Nomura $85 billion.
- T-Head founded in 2018 as Alibaba's chip arm.
- Products: Zhenwu AI chips, Yitian CPUs, Xuantie RISC-V licences.
- Over 60% of its compute serves outside customers (Alibaba).
- CEO Eddie Wu put annualised revenue around RMB10 billion.
9-point recap written from the linked reporting. The article itself is the source.
Chip Manufacturing
9 POINTS
Suzhou Denglin Technology, which designs general-purpose GPUs for AI workloads, began trading in Hong Kong on Tuesday after raising HK$1.41 billion in its initial public offering. It sold 6.125 million shares at HK$231 each, the bottom of its offer range, and expects HK$1.31 billion in net proceeds after expenses, according to its exchange filing. The listing came in a busy week for Chinese chip offerings in Hong Kong, with Alibaba's T-Head confirming its own filing the same day and SCMP reporting that one chip debut soared while another stumbled.
TechNode · November 10, 2026
Why it matters: China's GPU challengers are turning to public markets to fund the fight against Nvidia, and pricing at the bottom of the range shows investors are getting choosier.
Watchpoint: Denglin priced at the low end of its range; how the shares hold up after debut is the test of appetite for second-tier Chinese GPU makers.
- Began trading in Hong Kong on Tuesday, November 10.
- Raised HK$1.41 billion gross.
- 6.125 million shares sold at HK$231.
- HK$231 was the bottom of the offer range.
- Expected net proceeds: HK$1.31 billion.
- Develops general-purpose GPUs for AI workloads.
- Based in Suzhou.
- Figures from its HKEX filing.
- Same day Alibaba confirmed T-Head's IPO filing.
9-point recap written from the linked reporting. The article itself is the source.
Chip Manufacturing
9 POINTS
Fractile, a British AI-chip startup, raised a $720 million Series C led by Thrive Capital, Founders Fund, Lightspeed and Redpoint, with backing from the UK's Sovereign AI Fund. The British government put the valuation at 'over £5 billion'; the Financial Times reported about $7 billion. After a pivot, Fractile's design stacks ordinary DRAM directly on top of each chip to ease the trade-off between memory and speed that slows long agentic AI tasks, and it pairs the hardware with its own compiler. CEO Walter Goodwin says the money will let it ship its first chips and systems next year.
TechCrunch · November 11, 2026
Why it matters: Memory bandwidth, not raw compute, is the bottleneck for running large models, and a sovereign-backed British bet on fixing it is a rare European entry in the inference-chip race.
Watchpoint: Fractile has not yet shipped a chip; its first chips and systems are promised for 2027.
- Announced November 11.
- Series C: $720 million.
- Led by Thrive Capital, Founders Fund, Lightspeed and Redpoint.
- Backed by the UK's Sovereign AI Fund.
- UK government: valuation 'over £5 billion'.
- FT reported roughly $7 billion.
- Design stacks DRAM on top of each chip.
- Includes its own machine-learning compiler.
- First chips and systems due next year.
9-point recap written from the linked reporting. The article itself is the source.
Chip Manufacturing
9 POINTS
Helion, the Washington-state fusion-energy company, signed a letter of intent with the Commerce Department for up to $175 million through the CHIPS Research and Development Office to develop and manufacture high-voltage semiconductors for fusion power. With $50 million of private investment the package totals $225 million and will fund US semiconductor research, manufacturing and advanced-packaging capability. Helion's machines rely on pulsed electrical systems that must repeatedly switch and recover huge amounts of energy, and it says chips that perform at that level are not readily available. 'This award... will allow Helion to develop and manufacture advanced semiconductors here in the United States,' said CEO David Kirtley.
Semiconductor Digest · November 10, 2026
Why it matters: It extends CHIPS Act money beyond logic and memory into the power semiconductors that grid, fusion and data-center electrical systems all depend on.
Watchpoint: A letter of intent is not a final award; the funding still has to be finalised with Commerce.
- Announced November 10.
- Letter of intent with the Commerce Department.
- Up to $175 million via the CHIPS R&D Office.
- Plus $50 million of private investment, $225 million total.
- Purpose: high-voltage semiconductors for commercial fusion.
- Supports US research, manufacturing and advanced packaging.
- Helion is based in Washington state.
- Its pulsed-power systems need fast, high-energy switching.
- CEO David Kirtley calls it an investment in US chipmaking.
9-point recap written from the linked reporting. The article itself is the source.
Chip Manufacturing
9 POINTS
Infineon Technologies says it has processed 300mm silicon-carbide wafers into functional power devices at its fab in Villach, Austria, the first company to do so in a full-scale production environment, and did it using existing silicon manufacturing equipment. Larger wafers yield 2.3 times as many chips as today's 200mm SiC wafers, which promises lower costs for the power chips used in electric vehicles, renewable energy, industrial gear and AI data-center power supplies. Infineon will set the 300mm ramp timing with customers while continuing its 200mm SiC roadmap.
Semiconductor Digest · November 9, 2026
Why it matters: Silicon carbide is the efficient power chip inside EV chargers and AI server power supplies, and moving it to 300mm wafers is how its cost comes down.
Watchpoint: Infineon has given no date for volume production on 300mm SiC; it says ramp timing will be set with customers.
- Announced November 9 (US time).
- Functional power devices on 300mm SiC wafers.
- Made at Infineon's fab in Villach, Austria.
- Infineon says it is first in a full-scale fab environment.
- Uses existing silicon manufacturing equipment.
- 300mm wafers give 2.3 times more chips than 200mm.
- Markets: autos, renewables, industrial and AI data centers.
- Ramp timing to be planned with customers.
- The 200mm SiC roadmap continues in parallel.
9-point recap written from the linked reporting. The article itself is the source.
Chip Manufacturing
9 POINTS
TSMC said it remains focused on its logic foundry business after The Information reported, citing two people with direct knowledge, that the company had begun drawing up plans to manufacture 3D DRAM at Nvidia's request. A TSMC spokesperson told Tom's Hardware its mission is to supply logic chip technology and manufacturing capacity. 3D DRAM stacks memory cells vertically to raise density, one route to easing the memory shortage driven by AI hardware. The statement neither confirmed nor explicitly denied the reported planning.
TechNode · November 9, 2026
Why it matters: If the world's dominant foundry moved into memory it would upend the Samsung, SK Hynix and Micron business that AI accelerators depend on.
Watchpoint: TSMC's response restated its focus but did not directly deny that 3D DRAM plans were being drawn up.
- Reported November 9.
- The Information: TSMC drawing up 3D DRAM plans.
- The report says it was at Nvidia's request.
- Sourced to two people with direct knowledge.
- TSMC: its mission remains logic technology and capacity.
- Statement given to Tom's Hardware.
- 3D DRAM stacks memory cells vertically.
- Memory is a bottleneck for AI hardware.
- No explicit denial of the planning work.
9-point recap written from the linked reporting. The article itself is the source.
Chip Manufacturing
9 POINTS
AMD said on Tuesday that its sixth-generation Epyc SP7 server processor, with up to 256 cores, is generally available and shipping in partners' servers, the first of four 'Venice' family CPUs, with the other three due in 2027. The first Helion racks, built by hardware vendors from an AMD blueprint for agentic AI, become available through November. CEO Lisa Su says early demand for Venice is the strongest of any Epyc generation, with more than 240 platforms committed, 20% more than its predecessor. AMD claims SP7 systems deliver up to 3.7 times the performance of an Nvidia Vera CPU system.
Data Center Knowledge · November 10, 2026
Why it matters: Agentic AI runs a lot of ordinary CPU work around the GPUs, which is reopening the server-CPU market AMD and Nvidia now both want.
Watchpoint: The 3.7x and 2.2x comparisons with Nvidia Vera and Intel Xeon systems are AMD's own claims.
- Announced Tuesday, November 10.
- Epyc SP7: up to 256 cores, now generally available.
- First of four sixth-generation 'Venice' CPUs.
- The other three are expected in 2027.
- Helion agentic-AI racks available through November.
- More than 240 Venice platforms committed, 20% more than Turin.
- Lisa Su: strongest early demand of any Epyc generation.
- AMD claims up to 3.7x an Nvidia Vera CPU system.
- And up to 2.2x an Intel Xeon 6980P system.
9-point recap written from the linked reporting. The article itself is the source.
Chip Manufacturing
9 POINTS
Google Cloud expanded its eighth-generation Tensor Processing Unit family on Monday with two chips. TPU 8tx uses the same silicon as the TPU 8t training chip but adds hardware-level encryption and is paired with custom AMD CPU hosts so data and model weights stay protected even while inference runs. TPU 8i Max combines two TPU 8i inference chips in one package, doubling compute, SRAM and HBM bandwidth, with 384GB of high-bandwidth memory for agentic workloads with very large context windows. Google's Mark Lohmeyer said it is seeing 'tremendous demand' for TPUs.
Data Center Knowledge · November 9, 2026
Why it matters: Two new TPU variants in one year show how quickly the hyperscalers are tailoring their own silicon to specific AI jobs rather than buying general-purpose GPUs.
Watchpoint: Google gave no pricing or general-availability date for either chip in the announcement coverage.
- Announced Monday, November 9.
- Builds on the TPU 8t and 8i launched in April.
- TPU 8tx: same silicon as TPU 8t, plus hardware encryption.
- Paired with custom AMD CPU hosts.
- Protects data and weights during inference.
- TPU 8i Max: two TPU 8i chips in one package.
- Doubles compute, SRAM and HBM bandwidth.
- 384GB of high-bandwidth memory.
- IDC: specialised chips sharpen Google Cloud's edge.
9-point recap written from the linked reporting. The article itself is the source.
Chip Manufacturing
9 POINTS
Xilect, a robotics chip startup founded jointly by Chinese GPU maker MetaX and humanoid-robot maker UBTech Robotics, raised RMB 500 million in an angel round. It is designing a system-on-chip that combines multimodal perception, decision-making and motion control on the robot itself, with a software toolchain to let robot makers run existing AI models on it. The money will fund its first chip and software, a larger engineering team and testing with robot manufacturers. Xilect plans to give partners a development platform in 2027 and release its first chip in 2028.
TechNode · November 13, 2026
Why it matters: China's humanoid makers want their own robot brains rather than relying on Nvidia's Jetson and Thor modules, and this is a GPU maker and a robot maker building one together.
Watchpoint: Xilect's first chip is not due until 2028; for now it has funding and a design goal, not silicon.
- Reported November 13.
- Angel round: RMB 500 million.
- Co-founded by MetaX and UBTech Robotics.
- MetaX is a Chinese GPU maker.
- Designing a robot system-on-chip.
- Combines perception, decision-making and motion control.
- Toolchain to run existing AI models on it.
- Developer platform for partners in 2027.
- First chip planned for 2028.
9-point recap written from the linked reporting. The article itself is the source.
