NuAxi

Company and supply-chain hub

NVIDIA

NuAxi groups public briefings that mention NVIDIA: product launches, capacity and pricing signals, and supply-chain changes. The page updates continuously.

312 briefings trackedLatest briefings

Latest briefings

View more

Sat, Oct 31

  1. Impact score 60

    RTX 5080 Vanishes From NVIDIA's US Marketplace Two Weeks After the RTX 5090

    NVIDIA's US Marketplace no longer lists the GeForce RTX 5080; the GPU filter on the store's graphics card page now tops out at the RTX 5070 Ti, followed by the RTX 5070, RTX 5060 Ti, RTX 5060, and RTX 5050, and partner RTX 5080 models are gone as well. The same thing happened to the RTX 5090 about two weeks ago and, barring a brief cameo at $5,599, that card still hasn't returned; NVIDIA has denied that the RTX 5090 is discontinued, though it has not said anything about the RTX 5080 yet. Oddly, the RTX 5080 product page still says "Starting at $999." The removal seems limited to the US for now: NVIDIA's UK Marketplace still has 13 RTX 5080 listings and two RTX 5090 cards. The source does not state the cause, but notes a missing listing could simply mean there is no stock left, since the RTX 5080 already costs close to $2,000 at US retailers, roughly double its $999 MSRP. That leaves the RTX 5070 Ti as the fastest card on NVIDIA's US store, with seven listings ranging from $1,269.99 to $1,499.99, well above its $749 launch price.

    GeForce RTX 5080GeForce RTX 5090NVIDIATechPowerUpInventory Shift

Fri, Oct 30

  1. Impact score 55

    NVIDIA Reportedly Moves Up RTX 5080 24 GB Launch

    According to a post from Board Channels' ChannelGate account, NVIDIA is reportedly moving up the launch of the GeForce RTX 5080 24 GB, with availability now expected around January 2027 and December 2026 as the earliest option; earlier this month, various sources had relayed that the card was reportedly planned for Q1 2027. The same post says NVIDIA will stop supplying the RTX 5080 16 GB and the RTX 5090 D V2 to China, leaving the 24 GB RTX 5080 as the only flagship model in that market; the report only covers China, and there is no word yet on whether the 16 GB card is going away globally. The post also says NVIDIA has apparently already cleared out its RTX 5080 16 GB GPU inventory, with remaining stock expected to last around two months, roughly matching a leaker's claim that NVIDIA has January discontinuation dates set for the 16 GB card. If the January window pans out, a CES 2027 reveal seems likely, though NVIDIA has not confirmed anything. The post is machine-translated, so some wording may be off. Specs are still unknown, though 24 GB on the same 256-bit bus would most likely mean 3 GB GDDR7 modules instead of 2 GB ones. Board Channels previously suggested the card could cost around 19,999 yuan (nearly $3,000) in China, while the 16 GB model already sells for close to $2,000 at U.S. retailers, double its $999 MSRP. With the memory shortage expected to get worse in 2027, it is reasonable to expect the 24 GB card to be both expensive and hard to find. The source does not state the cause for the supply stop.

    GeForce RTX 5080 24 GBGeForce RTX 5080 16 GBGeForce RTX 5090 D V2NVIDIATechPowerUpEOL/Discontinuation
  2. Impact score 10

    Wafer-Level Burn-In Improves WBG Device Reliability, Automotive and Aerospace WBG Device Testing, Direct Dielectric Cooling for Power Converters: Power Electronics Week Insights

    Power Electronics News published a roundup of this week's must-read articles covering Wafer-Level Burn-In Improves WBG Device Reliability, Automotive and Aerospace WBG Device Testing, and Direct Dielectric Cooling for Power Converters. Items include: wafer-level burn-in screens SiC and GaN devices for infant mortality failures before packaging, improving module yields and reliability; direct dielectric cooling brings insulating liquid closer to SiC/GaN devices, shortening thermal paths but complicating insulation design; and WBG device test methodologies covering dynamic characterization, short-circuit ruggedness, thermal cycling, and gate oxide reliability. Other items: Infineon and NVIDIA advance GaN-based intermediate bus converters to improve power density and efficiency in 800 VDC AI data centers; Enphase and Infineon are integrating 900 V GaN bidirectional switches into SST systems designed for AI data center power conversion; Vertical Semiconductor targets $73M in proposed CHIPS funding to advance vertical GaN transistors aimed at improving AI data center power efficiency and density (proposed); Murata developed the world's smallest 008004-inch NTC thermistor, with the NCU01 series maintaining ±1% resistance and B-constant tolerances, initial samples planned for October 2026 and production in 2027; Flex completed the EPC Power acquisition, adding more than 15 GW of deployed power conversion technology and strengthening grid stabilization and backup power capabilities; SemiQ launched 750 V and 1200 V Gen4 SiC MOSFETs. The source is a news roundup and provides no price movement data.

    SiC devicesGaN devicesvertical GaN transistorsInfineon, NVIDIA and 8 totalPower Electronics News

Thu, Oct 29

  1. Impact score 60

    The high-capacitance #MLCC shortage persists! With #NVIDIA Vera Rubin ramping up and #AMD sourcing MI450 components, ODMs expect high-end MLCC shortag...

    TrendForce reported on October 29, 2026 that the high-capacitance MLCC shortage persists; with NVIDIA Vera Rubin ramping up and AMD sourcing MI450 components, ODMs expect high-end MLCC shortages to peak in Q1 and Q2 2027. TrendForce also noted that AI inference drives CPU and DDR5 demand, with high-capacitance MLCC shortages expected to persist.

    high-capacitance MLCChigh-end MLCCCPUNVIDIA, AMDTrendForce Official (Market Data)Inventory Shift
  2. Impact score 45

    Infineon Advances GaN Power for 800 VDC AI Data Centers

    Infineon Technologies AG is expanding its collaboration with NVIDIA on 800 V direct-current (VDC) power architectures for AI data centers, with the latest development focusing on high-voltage intermediate bus converters (IBCs), a stage in the power delivery chain that steps down the main distribution voltage to levels suitable for downstream conversion and processor power delivery. The collaboration builds on work announced in May 2025 and further detailed in October 2025, which addressed serviceable, hot-swappable server racks operating with 800 VDC power distribution. At the IBC stage, switching frequencies up to 1 MHz enable smaller magnetic components and passive devices, reducing converter size and board height; Infineon combines its GaN portfolio with PSoC™ Control C3 controllers and EiceDRIVER™ gate-driver products to provide integrated IBC solutions, supporting conversion ratios including 800 V to 50 V, 800 V to 12 V, and 800 V to 6 V. The collaboration with NVIDIA also extends beyond intermediate bus conversion to power-path protection, including hot-swap controller functionality intended to support safe servicing and the deployment of future server boards in 800 VDC architectures.

    GaN power deviceshigh-voltage intermediate bus converters (IBCs)PSoC™ Control C3 controllersInfineon Technologies AG, NVIDIAPower Electronics NewsPartnership/M&A
  3. Impact score 30

    Cloud to edge AI: If and when, to what degree, and how and why

    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.

    SSDDRAMHDDApple, Google and 6 totalEDN
  4. Impact score 48

    Bloomberg reports that #Marvell is in advanced talks to acquire optical communications component maker #Lumentum, with an announcement potentially com...

    Bloomberg reports that Marvell is in advanced talks to acquire optical communications component maker Lumentum, with an announcement potentially coming as soon as next week; whether the deal will be completed remains undetermined. Marvell is already a customer of Lumentum's laser products, while NVIDIA invested US$2 billion in each company earlier this year. TrendForce views the potential acquisition as a strategic move by Marvell to position itself early for optical interconnects at both the scale-in level (within servers and packages) and the scale-up level (within racks) in the CPO/NPO era. Marvell already offers electronic components such as DSPs and has gained optical I/O technology spanning package-to-rack connectivity through its acquisition of Celestial AI, while Lumentum supplies the indium phosphide (InP) lasers used as external light sources in NVIDIA's CPO switches. TrendForce notes that DSP functions are expected to gradually migrate into ASICs or electronic integrated circuits (EICs) as optical interconnects move deeper into racks and packages, while light sources remain difficult to integrate directly into silicon chips; with CPO penetration still low, the acquisition would represent a long-term positioning play.

    optical communications componentsindium phosphide (InP) lasersDSPsMarvell, Lumentum and 3 totalTrendForce Official (Market Data)Partnership/M&A
  5. Impact score 50

    RT NEWS: #Marvell is reportedly nearing a deal to acquire #Lumentum, possibly announced as soon as next week, Bloomberg says. After hours, Lumentum ro...

    According to Bloomberg, Marvell is reportedly nearing a deal to acquire Lumentum, possibly announced as soon as next week; this remains a reported/rumored deal that has not been confirmed or announced. In after-hours trading, Lumentum rose nearly 10% while Marvell fell over 7%. The text also mentions that NVIDIA invested $2B in each. The source does not state the cause.

    Marvell, Lumentum and 3 totalTrendForce Official (Market Data)Partnership/M&A
  6. Impact score 50

    RT NEWS: The two-year gaming #GPU upgrade cycle might be fading. #NVIDIA’s RTX 60 could reportedly slip to 2028, with #AMD’s RDNA 5 facing similar r...

    According to a news item reposted by TrendForce (posted 49 minutes ago), the two-year gaming GPU upgrade cycle might be fading. NVIDIA's RTX 60 could reportedly slip to 2028, with AMD's RDNA 5 facing similar rumors. The claim is reported as a rumor (reportedly/rumors), not an official confirmation. The source does not state the cause, only asking 'What's behind the slowdown?'.

    gaming GPUNVIDIA, AMDTrendForce Official (Market Data)
  7. Impact score 10

    [Tech Note] Why does AI still have to wait with tens of thousands of GPUs? How silicon photonics and CPO are building the ‘highways of light’

    This is a background technology column published on October 29, 2026 (authored by Professor Younghyun Kim of Hanyang University ERICA), explaining — not reporting — industry developments; it contains no price, capacity or policy event. The article states that AI data center bottlenecks have shifted from GPU compute speed to data movement: copper interconnects suffer signal loss, heat and repeater power overhead at high speeds and longer distances, so networks are shifting to silicon-photonics-based optical interconnects; co-packaged optics (CPO) places optical engines next to the switch ASIC to shorten electrical paths, increase bandwidth density and save power. It cites Meta's evaluation of a CPO switch system presented at ECOC 2025, in which power consumption in the optical interconnect portion was approximately 65% lower than a comparable pluggable architecture, with connection stability confirmed under the specific test conditions; companies such as NVIDIA and Broadcom are accelerating commercialization by introducing network products incorporating CPO technology. The article also describes LPO (removable modules with the DSP eliminated) and NPO (optical engines close to but outside the switch package) as alternative approaches, explains that HBM and CPO reduce GPU data-wait time at the memory and network levels respectively, and notes remaining CPO challenges including laser light supply, fiber connection, heat from ASICs, and yield improvement.

    co-packaged optics (CPO)silicon photonicsHBMNVIDIA, Broadcom and 3 totalSK Hynix Newsroom

Wed, Oct 28

  1. Impact score 60

    [Insights] NVIDIA RTX 60 May Slip to 2028, AMD RDNA 5 Also in Question: Why Are Gaming GPU Cycles Slowing?

    According to unconfirmed rumors, NVIDIA's RTX 60 series could be delayed until 2028, with similar speculation surrounding AMD's RDNA 5; if true, consumer GPU upgrade cycles could stretch to roughly three years. NVIDIA previously launched the RTX 30 series in 2020, the RTX 40 series in 2022, and the RTX 50 series in early 2025, maintaining a cycle of just over two years. The source states the cause: AI is a major factor — NVIDIA's August earnings report showed quarterly Data Center revenue of US$89 billion, approximately 92.5% of its US$96.2 billion total revenue; gaming is no longer its primary revenue driver, with R&D resources and memory supplies increasingly directed toward AI. Amid strong AI accelerator demand and rising memory prices, the RTX 50 SUPER series has reportedly been put on hold due to GDDR7 costs. Another factor is weakening upgrade incentive: TechSpot's testing found the RTX 5080 outperformed the RTX 4080 SUPER by an average of just 4% at 1440p and 11% at 4K in traditional gaming benchmarks without frame generation. Integrated graphics advances (Intel's Panther Lake Arc B390, AMD's Strix Halo Radeon 8060S, and the Blackwell GPU in RTX Spark) are also pressuring entry-level discrete GPUs. The source notes these performance results do not represent the entire RTX 50 lineup, and the delay reports remain conditional ('if true') speculation.

    consumer gaming GPURTX 60 seriesRDNA 5NVIDIA, AMDTrendForce
  2. Impact score 97

    NVIDIA GeForce RTX 5080 Reaches Nearly $2,000 Across U.S. Retailers

    NVIDIA's GeForce RTX 5080 has seen a significant increase in its listed price over the past few weeks, now costing nearly $2,000 at U.S. retailers. Cause: price hikes throughout the supply chain and the GPU's 16 GB of GDDR7 memory, which is currently in short supply, making the memory modules much more expensive. The card launched in early 2025 with an MSRP of $999; add-in card partners' custom solutions (new PCBs, enhanced cooling, and other features) gradually increased the price, and the issue escalated when memory supply became constrained. Across U.S.-based retailers like Best Buy, Micro Center, Amazon, Newegg, and B&H, the listing price now starts at nearly $2,000; on B&H, older listings show better pricing but are marked 'Temporary out of stock,' with only in-stock listings priced at $2,000 or more. In two months, listings have jumped from $1,200 to $2,000.

    GeForce RTX 5080 GPUGDDR7 memoryNVIDIATechPowerUpPrice Movement
  3. Impact score 60

    TSMC Could Charge Up to 20% More for CoWoS Packaging Amid AI Crunch

    According to supply chain sources cited by Taiwan's Commercial Times, TSMC is reportedly considering raising prices for its CoWoS (Chip on Wafer on Substrate) advanced packaging by as much as 20%; TSMC has not confirmed anything yet. Foundries and outsourced packaging firms are apparently weighing another round of 10% to 20% increases, as CoWoS demand has spread beyond AI GPUs to server CPUs and custom ASICs, leaving a growing backlog of unfilled orders. Testing is turning into a bottleneck of its own: Nomura estimates, via Seoul Economic Daily, that NVIDIA's Blackwell takes roughly four times longer than Hopper in final testing, while Rubin could need seven times as long. Packages keep growing, with TSMC already mass producing CoWoS-L at 5.5 times the reticle limit. Tool lead times have reportedly stretched from three to six months to as long as eight months, and TSMC, which usually negotiates for discounts or extra features, apparently placed this year's equipment orders at existing prices.

    CoWoSCoWoS-LTSMC, NVIDIATechPowerUpPrice Movement
  4. Impact score 50

    #Google’s #TPU push is gaining momentum as #OpenAI reportedly eyes its AI chips and #Oracle plans to offer TPUs through its cloud. Could this challen...

    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
  5. Impact score 55

    Samsung 8LPU+ Gets Floated as the Node for a Future Switch 2 Lite

    Samsung Foundry is scheduled to present 8LPU+, the second generation of its 8LPU platform, at IEDM 2026 on December 14. Nintendo Patents Watch has already pegged it as a possible fit for a Switch 2 Lite, but this is speculation for now: Nintendo has not announced a Switch 2 Lite, and Samsung's material does not mention Nintendo or NVIDIA. The link exists because the Switch 2's NVIDIA T239 SoC is built on Samsung's 8 nm node, so a refined version of that process is an obvious candidate for a cheaper handheld-only model. Samsung's IEDM abstract lists 12% higher performance, 15% lower power, and an 8% area shrink over 8LPU, along with roughly 10% smaller I/O area and lower standby leakage. It is a DUV-only node, and Samsung pitches it at automotive, robotics and edge AI chips. Korea's Electronic Times reports the platform should arrive before year's end, and quoted an industry official who named existing 8 nm customers as likely candidates. These are process-level figures, so they do not directly translate into a future Switch 2 Lite that runs 12% faster or lasts 15% longer per charge, and it is unknown how closely NVIDIA's custom 8N variant maps onto 8LPU+.

    Samsung 8LPU+Samsung 8LPUSamsung 8 nm nodeSamsung Foundry, Nintendo and 3 totalTechPowerUpCapacity Change
  6. Impact score 60

    Samsung Electro-Mechanics and NVIDIA Ink ABF Substrate Supply Agreement

    Samsung Electro-Mechanics and NVIDIA have signed an ABF substrate supply agreement. The companies will expand dedicated production capacity in South Korea to supply substrates for AI processors and high-speed networking switches. The source does not state the cause, and provides no specific dates, capacity figures or investment amounts; the capacity expansion is phrased as a plan ('will expand').

    ABF substrateSamsung Electro-Mechanics, NVIDIAEE Times AsiaPartnership/M&A
  7. Impact score 55

    Zotac Restocks NVIDIA GeForce RTX 5090 Solid OC in Japan Amid Reports of GPU's EOL Status

    The latest development in the rumors of an NVIDIA GeForce RTX 5090 discontinuation is that OEMs are reportedly no longer able to order GPUs or VRAM for the 5090 from NVIDIA; however, Zotac appears undeterred, having just published a restock of the GeForce RTX 5090 Solid OC in Japan. The new listing appeared on Zotac's official online store in Japan and was spotted by @momomo_us on X, a popular leaker and hardware insider. While the GPU appears to be in stock and selling for as much as 1,190,000 JPY (~$7,558.17 converted), there is no telling whether this will remain the case. Zotac may have had some stock ready to go, and the previously reported lack of supply or back-order availability from other board partners could instead have been due to regional or AIB limitations on GPU allocation by NVIDIA. It may also indicate that previous reports of the RTX 5090 being end-of-life were inaccurate.

    NVIDIA GeForce RTX 5090VRAMZotac, NVIDIATechPowerUpEOL/Discontinuation

Tue, Oct 27

  1. Impact score 50

    Is #NVIDIA easing the HBM pressure? Beyond rethinking Rubin Ultra’s memory mix, the chipmaker has quietly cut Rubin GPU’s HBM4 bandwidth by nearly 1...

    According to a TrendForce post dated Oct 27, 2026 (reported, unconfirmed), NVIDIA is reportedly reconsidering 12-Hi HBM4E for Rubin Ultra in favor of a mix of 8-Hi HBM4 and HBM4E, which could potentially ease costs and supply constraints. Separately, the chipmaker has quietly cut Rubin GPU's HBM4 bandwidth by nearly 13%. Both items remain reported/conditional claims and are not confirmed as implemented.

    HBM4EHBM4NVIDIATrendForce Official (Market Data)
  2. Impact score 60

    [News] Apple iPhone Duo Faces Reported Hinge Defects; Shipment Target Seen Up 44% on Strong Demand

    Apple’s first foldable smartphone, the iPhone Duo, is reportedly facing quality concerns over display damage near the hinge. According to MacRumors on October 26, Apple acknowledged an issue affecting the inner displays of a small number of iPhone Duo units, but did not disclose the cause of the cracking or whether a fix is planned; affected customers have been able to obtain in-store replacements, though limited inventory has left some waiting. According to ZDNet (citing MacRumors), persistent hinge problems during development prompted supplier changes: U.S.-based Amphenol was initially the primary hinge supplier with Taiwan’s Shin Zu Shing as secondary; Shin Zu Shing took over after Amphenol’s production yields fell well below expectations, but reportedly also struggled, having to select units meeting Apple’s quality standards as “golden samples,” and these difficulties delayed iPhone Duo production. Reuters adds that iFixit, via a 3D hinge simulation based on teardown footage, patents, and CT scans, suggested a possible cause: a central support bar may fail to retract properly during folding, placing pressure on the inner display. According to NewsPim, Apple has reportedly asked Foxconn and component suppliers to raise the iPhone Duo shipment target for the next 12 months by 44%, from 9.7 million to 14 million units; industry sources suggest the increase reflects stronger-than-expected initial demand plus additional component stockpiling for potential assembly defects, and Foxconn’s final assembly yield was reportedly around 60% in September, with hinge-related defects potentially further increasing demand for replacement OLED panels. TrendForce notes early screen-failure reports could undermine consumer confidence, with the long-term impact largely depending on how widespread the issue is and how quickly Apple identifies the root cause and repairs or replaces affected units. Separately, per ZDNet, another potential supply constraint is T-glass, a high-end fiberglass yarn used in copper-clad laminates (CCLs) for PCBs, where Japan’s Nittobo holds a near-monopoly and is prioritizing shipments for AI chips made by companies such as NVIDIA, potentially adding pressure to Apple’s supply chain.

    hingeinner OLED displayT-glassApple, Amphenol and 6 totalTrendForce
  3. Impact score 60

    [News] NVIDIA’s Memory Shake-Up: Rubin Ultra Rethinks 12-Hi HBM4E for 8-Hi HBM4, Rubin GPU Bandwidth Cut 13%

    According to ZDNet, Samsung and SK hynix are preparing to ship 8-Hi HBM4 for Rubin Ultra in 2Q27, departing from the original 12-Hi HBM4E plan; 8-Hi HBM4 was initially absent from NVIDIA's plans, but talks gained traction in 3Q26. The source cites thermal challenges and lower yields in 12-Hi stacks, potentially easing supply constraints. Although both Samsung and SK hynix have completed NVIDIA's qualification for 12-Hi HBM4, their 8-Hi products require separate certification; a source said NVIDIA is expected to deploy both 8-Hi HBM4 and 8-Hi HBM4E in Rubin Ultra-based NVL rack-scale systems depending on rack configuration. TrendForce says 8-Hi configurations use fewer DRAM dies, lowering cost per stack and GPU bill of materials, and given tight HBM supply, 8-Hi setups are expected to be a priority for evaluation by many GPU and ASIC vendors in 2027; ZDNet reports 12-Hi HBM4 remains mainstream in 2026, with 8-Hi adoption expected to rise sharply from mid-2027. Economy Tribune, citing Daishin Securities, suggests Anthropic's in-house AI chip could also feature 8-Hi HBM4, raising the prospect of Samsung supplying both foundry services and memory. Separately, per Wccftech, NVIDIA trimmed Vera Rubin's HBM4 bandwidth specifications: an archived September 1 version listed the Vera Rubin NVL72 system with 20.7 TB of HBM4 capacity and 1,580 TB/s bandwidth, the Vera Rubin Superchip at 576 GB and 44 TB/s, and each Rubin GPU at 288 GB and 22 TB/s; NVIDIA has since lowered per-GPU HBM4 bandwidth to 19.2 TB/s and NVLink 6 throughput from the originally advertised 3.6 TB/s to 3 TB/s, cuts of roughly 13% and 17% respectively, and up to 26% for HGX Rubin NVL8 systems, with changes dating back to September. IT Home notes NVIDIA revised the Vera CPU memory specification from a fixed 1.5 TB of LPDDR5X per chip to 'up to 1.5 TB', and that lowering bandwidth requirements could allow more HBM4 products to qualify for Rubin GPUs, potentially boosting overall production output.

    HBM4HBM4ELPDDR5XNVIDIA, Samsung and 4 totalTrendForceCapacity Change