NuAxi groups public briefings that mention Google: product launches, capacity and pricing signals, and supply-chain changes. The page updates continuously.
Google is deepening its strategic partnership with Foxconn, aiming to cut the time from chip supply to data-center deployment from the traditional 12 to 18 months to about six months. The statement was made by Rehan Sheikh, vice president of global silicon chip technology and manufacturing at Google Cloud, at Hon Hai Tech Day (HHTD 26) on October 30. The six-month timeline is a stated aim, not an achieved result; the source does not state the cause.
This is a weekly industry roundup listing multiple semiconductor developments: HBM shortages threaten AI expansion; TSMC weighs price hikes (under consideration only, not confirmed); advanced packaging ramps; domain-specialized AI models; Infineon and UMC boost fab investments; Intel's quantum move; Lam and imec advance 3D integration; Google expands TPUs; AI-driven earnings; and CPU benchmarking methodology for agentic AI. The source provides only headline-level mentions, with no specific figures, dates or causes, and does not state implementation status of these items.
HBMTSMC, Infineon and 7 totalSemiconductor Engineering
Anthropic's "multi-platform resource pool" strategy, disclosed in its S-1 filing, is sending an important signal to the AI chip industry. The company uses Amazon Trainium, Google tensor processing units (TPUs), Nvidia GPUs, and AMD GPUs, integrating computing capacity from different partners into a unified resource pool through software and allocating it across chip platforms based on task requirements. The source states this strategy pressures Nvidia's pricing power.
AI chipsNvidia GPUsAMD GPUsAnthropic, Nvidia and 5 totalDigiTimes Asia
Reported on 2026-10-30, Intel's 3Q26 results were mixed: Data Center and AI (DCAI) revenue surged 78% to $7.3 billion and operating income more than tripled to $3.04 billion; however, according to Reuters, a fatal explosion in October at Intel's Kiryat Gat fab in Israel, which primarily produces data center processors using 10nm technology, is expected to cost roughly $600 million in revenue over two quarters, and sluggish PC demand could persist through 1H27. Reflecting these headwinds, Intel forecast fourth-quarter revenue of $15 billion to $16 billion, missing the $17.07 billion LSEG consensus, with adjusted earnings projected at 25 cents per share versus analysts' 43-cent expectation. CFO David Zinsner said the incident would reduce fourth-quarter revenue by roughly $300 million and shave 200 basis points off gross margin, with a similar revenue hit expected in 1Q27; after a two-week shutdown, Intel aims to begin restoring production at the Israeli facility by the end of this week. In the PC business, 3Q26 sales fell 5% to $8.10 billion, missing analysts' $8.98 billion estimate, and Zinsner expects fourth-quarter PC business revenue to plunge by a percentage in the mid-20s. CEO Lip-Bu Tan said he was disappointed the company could not meet strong customer demand, citing manufacturing supply shortages, an unprecedented memory crunch and resulting PC demand weakness as near-term challenges; he also noted that despite rapidly ramping Intel 3 wafer output, an industry-wide shortage of advanced organic substrates for EMIB packaging continues to limit production scaling. Zinsner also warned the chipmaker is unlikely to generate positive cash flow in 2026; Intel had raised its capital expenditure guidance from $18 billion to $20 billion in July, with a further substantial increase expected in 2027 to support its manufacturing buildout. On the positive side, Zinsner expects industry-wide server CPU unit shipments to grow more than 30% in both 2026 and 2027; Intel Foundry revenue surged 58% to $6.7 billion in 3Q26 including internal sales, with its operating loss narrowing to $1.85 billion from $2.32 billion a year earlier. Tan said a couple of customers will be ready to announce adoption of Intel's EMIB/EMIB-T advanced packaging over the next two quarters, and that in October Intel delivered its industry-standard 0.9 process design kit (PDK) for the 14A process to select customers as planned. TrendForce expects Google to adopt EMIB-T in 2027, while AWS is also testing EMIB technology; industry sources earlier suggested Google's TPU v9 lineup includes three variants: Broadcom-designed WhaleFish (expected to use TSMC's CoWoS-L with mass production slated for 2H27), MediaTek-designed HumuFish (set for 1H28 mass production relying exclusively on Intel's EMIB-T) and TriggerFish (also planned to use EMIB-T, targeted for 2H28 or later).
EMIBEMIB-Tadvanced organic substratesIntel, Google and 5 totalTrendForceDisaster/Disruption
This is an opinion/analysis article arguing that AI processing, specifically inference, remains mostly cloud-centric but is shifting toward edge devices, and that storage and memory requirements will keep rising in both cloud and edge scenarios. According to the article, data center build-outs are redirecting supplies of SSDs (and HDDs) and DRAM, while the required capacities in PCs, smartphones and the like for today's edge AI processing are increasing in lockstep: the Gemini AI models embedded in latest-generation Chrome browser installs take up at least 4 GBytes of storage space; Apple Intelligence included with iOS 27 is now required versus optional with the version 26 predecessor and uses up to 14 GBytes; the required-by-default AI model suite included with MacOS 27 is 12-30 GBytes in size. PrismML compressed Alibaba's Qwen3.8 27B open-source model by 9-10x while still matching 98% of Qwen's aggregate benchmark scores, but the resultant model was still 5.9 GBytes; Perplexity's Portable Computer for local on-computer model storage and processing uses 20 GBytes. The article also states that the training run for OpenAI's latest-generation "Astra" GPT-6 fully consumed more than 100,000 GPUs, and argues that hybrid approaches (local base-level inference with cloud tapped as needed) will become increasingly common, with the decode portion of inference being memory-intensive from both capacity and bandwidth standpoints, rationalizing advanced HBM approaches. All points are the author's analysis; the source reports no price changes.
Google is now supplying its tensor processing units (TPUs) to companies it competes with, including Anthropic and Apple. On Alphabet's third-quarter earnings call, management gave two reasons: outside customers keep the chips competitive, and outside volume lowers the cost of every chip Google uses itself. However, it would not say how much money the business makes, and the source does not state any specific revenue or profit figure for the business.
tensor processing units (TPUs)GoogleDigiTimes Asia
Google's parent, Alphabet, is turning demand for artificial intelligence into a larger cloud business. Its latest quarter shows that stronger operating performance does not yet remove the financial burden of building the infrastructure behind it: cloud revenue more than doubled, while spending on property and equipment exceeded the cash generated by operations. The source provides no specific amounts, percentages, or quarter start and end dates.
TrendForce posted at 12:00 PM on Oct 28, 2026 that Google's TPU push is gaining momentum: OpenAI reportedly plans to buy Google TPUs, and Oracle plans to bring TPUs to its cloud in 2027. TrendForce raises the question of whether this could challenge NVIDIA's dominance. Both items remain reported plans/rumors rather than confirmed or implemented transactions. The source does not state the cause.
TPUGoogle, OpenAI and 4 totalTrendForce Official (Market Data)Partnership/M&A
Oracle has signed partnerships with Cerebras and other chipmakers to host their chips in its cloud infrastructure. The Cerebras deal gives the chipmaker a new large customer to anchor its post-IPO trajectory, while allowing Oracle to reduce its reliance on Nvidia's expensive chips; even so, Oracle also announced a purchase of Vera Rubin platforms for its data centers.
AI chipsCerebras chipsVera Rubin platformsOracle, Cerebras and 5 totalDigiTimes AsiaPartnership/M&A
Google's tensor processing unit (TPU) business has gained one confirmed and one reported major customer within a week: Oracle is the confirmed customer and OpenAI is reportedly next. This adds pressure on Nvidia, whose graphics processors still dominate AI computing even as its biggest buyers look for alternatives.
Google TPUNvidia graphics processorsGoogle, Nvidia and 4 totalDigiTimes Asia
Google recently unveiled its flagship AI model, Gemini 4 Argon. The model consumed an unprecedented 25 trillion tokens within 72 hours of launch—three times the volume recorded by Gemini 3.0 over the same timeframe, setting a new record for the company. In his first in-depth public interview since the release, Google CEO Sundar Pichai addressed market skepticism regarding claims that 'Google has abandoned frontier AI' and outlined Google's strategy for allocating AI compute resources, managing its cloud business, and deploying long-term infrastructure. The source title indicates that Google prioritizes frontier AI as cloud capacity runs short.
AI compute capacitycloud capacityGoogleDigiTimes Asia
Semiconductor IP company Chips&Media disclosed its third-quarter 2026 results on Oct. 27, reporting consolidated revenue of 7.162 billion won, up 8.2% year on year, and operating profit of 1.657 billion won, up 8.7% year on year. Growth was driven by IP licensing revenue: third-quarter licensing revenue rose 33% year on year to 4.431 billion won. The company secured its first licensing contract related to a domestic on-device AI semiconductor project, won a new automotive project from a major South Korean fabless chipmaker, signed new licensing agreements with Chinese customers for AI SoCs and humanoid robots, and signed a new mobile licensing agreement with existing customer Google. Royalty revenue, however, fell 26.5% year on year to 2.165 billion won; per the source, the decline reflected reduced chip production by customers amid memory supply shortages and rising prices. The company expects royalty revenue to return to growth in 2027, with domestic fabless AI SoC customers scheduled to begin mass production of chips for robots, home appliances and IoT devices. Service revenue increased 66.8% to 565 million won, and the company swung to a net loss of 605 million won, which the source attributes to foreign exchange losses and valuation losses on financial assets. Chips&Media also said it plans to strengthen its position in the mobility market, expand into edge AI fields including robotics and drones, address demand for next-generation video codec standards APV and AV2, commercialize AI ISPs and further enhance its NPU IP.
IP licensingRoyalty revenueService revenueChips&Media, GoogleThe Elec
At its Las Vegas event, Oracle announced an expanded partnership with Google and plans to bring Google's Tensor Processing Units (TPUs) to Oracle Cloud Infrastructure (OCI), with the chips available for selection within OCI beginning in 2027 and capacity increasing over the course of several years; Oracle says the deal will make it the first hyperscale cloud platform outside Google's own cloud services to offer the chips, though it did not mention the quantity of TPUs to be available next year or beyond. OCI president Mahesh Thiagarajan said different AI workloads have different compute needs and that bringing Google TPUs to OCI gives customers the flexibility to choose the right infrastructure for the performance, throughput, and economics of each AI workload. Alongside the TPU plan, Oracle announced on Tuesday the general availability of Nvidia Vera Rubin NVL72 systems and Nvidia Vera CPUs, plus plans to introduce HGX Rubin NVL8 systems without specifying when those would become available, as well as new ChatGPT integrations, Oracle AI Workspace, Oracle AI Security Mesh, and a deal with Snowflake. Oracle also posted record earnings in the quarter ending August 31, with revenue up 30 percent year over year.
Google Tensor Processing Units (TPUs)Nvidia Vera Rubin NVL72 systemsNvidia Vera CPUsOracle, Google and 3 totalThe RegisterPartnership/M&A
Anthropic's IPO filing reveals an infrastructure strategy extending far beyond conventional cloud procurement: the Claude developer has locked in hundreds of billions of dollars of compute capacity across Google, Amazon, Microsoft, Broadcom, and a growing roster of GPU infrastructure providers, spanning TPUs, GPUs, and dedicated data centers. This is a capacity commitment arrangement disclosed in the IPO filing; the source does not state the cause.
AWS has launched the open-source TorchNeuron tool, allowing developers to port PyTorch artificial intelligence (AI) applications to its custom Trainium chips without rewriting code. Google has followed with TorchTPU, while AMD and Intel are integrating AI software optimizations, as chipmakers challenge Nvidia's market advantage built on the CUDA ecosystem.
TrainiumTorchNeuronTorchTPUAWS, Google and 5 totalDigiTimes Asia
Google's Project Suncatcher is testing AI hardware in space; in that environment, radiation, cooling, power, and hardware lifetime determine the economics of computing. The article is titled "In Orbit, Useful Life Becomes a Performance Metric," and the source states no further details beyond these factors, with no specific part numbers, quantities, or dates given.
This is a roundup of the site's most popular articles from the last week. No. 5: Google and Synaptics have introduced the xMCU – an MCU combining heterogeneous, AI-native compute and acceleration with integrated high-bandwidth custom DRAM scalable to hundreds of megabytes; the Astra SR200 Series targets AI wearables and battery-powered sensing devices, enabling larger AI models and richer multimodal workloads entirely on-device. No. 4: According to the Data Centers Processor Market Monitor, Q3 2026, agentic AI is reviving datacenter CPU demand, with CPU revenues up 35%, ahead of AI ASICs (+23%) and GPUs (+13%); GPU and AI ASIC designers are scaling back their roadmaps, sometimes drastically, in response to the sharp rise in DRAM prices. No. 3: London neocloud startup Nscale has had another setback to its IPO plans as Microsoft pulled out of a deal to be the anchor tenant on Nscale's Loughton, Essex, datacentre; neither party gave any reason for the pull-out, but UK Power Networks has told Nscale it cannot supply enough power for the site until the early to mid 2030s. No. 2: Per the Seoul Economic Daily, Nvidia has asked Samsung, Hynix and Micron to use base dies made on TSMC's 3nm node for its custom HBM called NVHBM; the base die contains the logic layer at the bottom of the HBM stack that handles the memory controller and interface, and the source's report of Nvidia's reason is truncated. No. 1: According to a memo from CEO Cristiano Amon to employees seen by Reuters, Qualcomm is to lay off staff worldwide, with notifications to start next month; the number involved is not known, the company had 52k employees in September 2025, and the memo describes the lay-offs as "a performance-focused reduction after annual reviews".
xMCUNVHBM base dieHBMGoogle, Synaptics and 10 totalElectronics Weekly
According to industry sources (reportedly, unconfirmed), Google is reportedly sticking with Intel EMIB-T despite substrate yields of only 35%, as the HumuFish variation's schedule tightens. Industry sources point to the HumuFish variation as a key area of concern, with reported yield problems. Uncertainty is growing over potential delays to Google's TPU v9; the source does not state that any delay has occurred and gives no specific timeline, with the reported yield problems serving as the stated cause of concern.
TPU v9HumuFishEMIB-T substrateGoogle, IntelTrendForce Official (Market Data)
At PyTorch Conference 2026, held in San Jose, California, on Oct. 20 local time, Google and Amazon unveiled the open-source frameworks TorchTPU and TorchNeuron, respectively, designed to simplify migrating existing AI applications from Nvidia GPUs to their own chips and to challenge Nvidia's CUDA ecosystem dominance. PyTorch, developed in 2016 primarily by Meta's (formerly Facebook's) AI research lab FAIR and used by approximately 85% of AI application developers worldwide, was originally designed around Nvidia GPUs and the CUDA platform, making migration to other chips difficult. TorchTPU, developed by Google in collaboration with Meta, according to Google, allows more than 98% of existing PyTorch models to run on TPUs simply by changing the device string in the code to model.to("tpu"), with conversion taking from a few minutes to a few days, and Google said it enabled TPUs to operate as native devices on an equal footing with GPUs and CPUs at the core PyTorch level. AWS's TorchNeuron similarly allows PyTorch-based AI models to run directly on AWS's Trainium AI chips, requiring developers to change only a few lines of device settings, replacing Nvidia's cuda with AWS's neuron. Amazon has released TorchNeuron as a public beta through GitHub under the Apache 2.0 open-source license and is also contributing code directly to the main PyTorch open-source project. Previously, migrating AI applications optimized for Nvidia's ecosystem to Google's or Amazon's proprietary AI chips took at least several days and, in more complex cases, several months; the new tools effectively remove these time and cost barriers. As TorchNeuron, TorchTPU and similar tools gain adoption, dedicated AI chips are expected to compete primarily on underlying performance, power efficiency and pricing, though additional time will be needed to verify compatibility, performance and costs for individual models.
Nvidia CUDATorchTPUTorchNeuronGoogle, Amazon and 5 totalThe Elec
Related manufacturers
Google, Amazon, Meta, Nvidia, Amazon Web Services (AWS)
Google reportedly remains committed to Intel's EMIB-T for HumuFish despite poor substrate yields. According to industry sources, EMIB-T substrate yields were around 35% in 3Q26, well short of the 50% target set for year-end; the reported causes are thermal expansion differences between ABF substrates and embedded silicon bridges that make precise alignment difficult during high-temperature processing, and NCF (Non-Conductive Film) applied beneath the bridges' TSVs trapping air bubbles during bonding, potentially weakening adhesion. Despite market rumors of a HumuFish delay and the possibility of a switch to TSMC's CoWoS-L, supply chain checks indicate HumuFish remains '100%' committed to EMIB-T. Intel has adjusted its substrate electroplating process, and suppliers including Ibiden, Unimicron, and Shinko continue working to improve yields. Google's TPU v9 lineup is confirmed to include three variants: Broadcom-designed WhaleFish (expected to adopt TSMC's CoWoS-L, with mass production slated for 2H27), MediaTek-designed HumuFish (set for 1H28 mass production, relying exclusively on Intel's EMIB-T), and TriggerFish (also planned to use EMIB-T, with production targeted for 2H28 or later). Per TechNews citing analyst Ming-Chi Kuo, TriggerFish is positioned as an upgraded TPU v9 with stronger inference capabilities and features three key upgrades over HumuFish: 2–3 times the SRAM capacity, an additional simulation die, and a shift to HBM4E from HumuFish's HBM4. On delays, the tape-out of Google-designed Compute Die A, built on TSMC's N2 process, has been pushed back to late October while other dies remain on schedule, and the delay is expected to push back trial production of the complete package; Commercial Times further notes the MediaTek-involved Google TPU v9 project (HumuFish) could see its tape-out pushed back to January 2027 due to design revisions, potentially delaying shipments until 1Q28. With HumuFish not slated for mass production until 1H28, the supply chain still has time to address EMIB-T's remaining issues, but a tape-out slip to January 2027 could compress that cushion.
Intel EMIB-T substratesABF substratesNCF (Non-Conductive Film)Google, Intel and 8 totalTrendForce