Valve announced three new hardware products: Steam Frame, Steam Machine and Steam Controller. The Steam Frame is Valve’s newest VR headset, while the Steam Machine is a PC gaming setup for the living room. As for the Steam Controller, it is marketed to replace the user’s mouse. The PC home console is specifically designed with the Steam Controller in mind.

Here are the tech specs for the Steam Machine, Valve’s take on a home console, according to an official posting:

General

• CPU
  • Semi-custom AMD Zen 4 6C/12T
  • Up to 4.8 GHz, 30 W TDP
• GPU
  • Semi-Custom AMD RDNA3 28CUs
  • 2.45 GHz max sustained clock, 110 W TDP
• RAM
  • 16 GB DDR5 + 8 GB GDDR6 VRAM1
• Power
  • Internal power supply, AC power 110-240V2
• Storage
  • Two Steam Machine models
    • 512 GB NVMe SSD3
    • 2 TB NVMe SSD
  • Both models include a high-speed microSD card slot

Connectivity

• WiFi
  • 2×2 WiFi 6E
• Bluetooth
  • Bluetooth 5.3 dedicated antenna
• Steam Controller
  • Integrated 2.4 GHz Steam Controller wireless adapter

I/O

• Displays
  • DisplayPort 1.44
    • Up to 4K @ 240Hz or 8K@60Hz5
    • Supports HDR, FreeSync and daisy-chaining
  • HDMI 2.0
    • Up to 4K @ 120 Hz
    • Supports HDR, FreeSync and CEC
• USB
  • Two USB-A 3.2 Gen 1 ports in the front
  • Two USB-A 2.0 high-speed ports in the back
  • One USB-C 3.2 Gen 2 port in the back
• Networking
  • Gigabit ethernet7
• LED Strip
  • 17 individually addressable RGB LEDs for system status and customizability

Size and Weight

• Size
  • 152 mm tall (148 mm without feet), 162.4 mm deep, 156 mm wide
• Weight
  • 2.6 kg

Software

• Operating System
  • SteamOS 3 (Arch-based)
• Desktop
  • KDE Plasma

Here are the tech specs for the controller, according to Valve’s official posting: 

Controls and Input

• Gamepad controls
  • A, B, X, Y buttons
  • D-pad
  • L and R analog triggers
  • L and R bumpers
  • View and Menu buttons
  • Steam and QAM buttons
  • 4x assignable grip buttons
• Thumbsticks
  • 2x full-size magnetic thumbsticks (TMR) with capacitive touch
• Haptics
  • 4x haptic motors
  • 2x LRA haptic motors in trackpads for HD tactile feedback
  • 2x high-output LRA haptic motors in grips for HD game haptics, including rumble
• Trackpads
  • 2x 34.5mm square trackpads with haptic feedback
  • Pressure-sensitive for configurable click strength
• Gyro
  • 6-axis IMU8
• Grip Sense
  • 2x capacitive areas along back of Steam Controller handles

Connectivity

• Steam Controller Puck
  • 2.4 GHz Wireless connection
  • ~8 ms full end-to-end, 4 ms polling rate (measured at 5m)
  • Up to 4 Steam Controllers per Steam Controller Puck
  • Steam Controller Puck connects to PC via USB-C
• Bluetooth
  • Bluetooth 4.2 minimum, 5.0 or higher recommended
• USB
  • USB-C tethered play

Power

• Charging
  • Steam Controller Puck charging interface
  • USB-C connector
• Battery
  • 8.39 Wh Li-ion battery
  • 35+ hours of gameplay
  • Battery life for tracked gameplay with Steam Frame is reduced

Size and Weight

• Size
  • Steam Controller: 111mm x 159mm x 57mm
  • Steam Controller Puck: 50mm x 28mm x 9mm
• Weight
  • Steam Controller: 292 g
  • Steam Controller Puck: 16 g

As for Steam Frame, the VR headset is marketed as a VR and non-VR headset. It is “designed for comfort” and ease of use; it comes with a separate set of controllers. The hardware’s production was leaked earlier, having been codenamed “Deckard.”

Here are the Steam Frame tech specs, according to the official marketing post:

General

• Processor
  • 4 nm Snapdragon® 8 Gen 3
  • Architecture: ARM64
• RAM
  • 16 GB Unified LPDDR5X RAM
• Storage
  • 256 GB/1 TB UFS storage options
  • microSD card slot for expanded storage
• Power
  • Rechargeable 21.6 Wh Li-ion battery
  • One USB-C 2.0 port in the rear, for charging and data
  • Charge with USB-C, 45 W
• Modular Headstrap
  • The headstrap includes integrated dual audio drivers and and rechargeable battery on rear.
  • Headstrap weight: 245 g
  • The core module can be separated from the headstrap for other headstrap solutions.

Display and Optics

• Display
  • 2160 x 2170 LCD (per eye)
  • 72-144 Hz refresh rate (144 Hz experimental)
• Optics
  • Custom pancake lenses
  • Glass and non-glass optical elements
  • Large FOV (up to 110 degrees)
  • IPD target range
    • 60-70 mm
  • Eyeglasses max width
    • 140 mm

Cameras and Tracking

• Tracking
  • Inside-out camera-based tracking
• Cameras
  • 4x outward-facing monochrome cameras for controller and headset tracking
  • 2x interior cameras for eye tracking and foveated streaming
• Passthrough
  • Monochrome passthrough via outward-facing cameras
• Low-light support
  • IR illuminators for tracking and passthrough in dark environments
• Expansion
  • User-accessible front expansion port
  • Dual high-speed camera interface (8 lanes @ 2.5 Gbps MIPI) / PCIe Gen 4 interface (1-lane)

Connectivity

• WiFi
  • WiFi 7, 2×2
  • Dual radios enable concurrent 5 Ghz Wi-Fi and 6 Ghz VR streaming
• Wireless Adapter
  • Wireless adapter included in the box
  • WiFi 6E (6 GHz)
  • Provides direct, low-latency link between headset and PC
• Bluetooth
  • Bluetooth 5.3

Audio

• Speakers
  • Dual speaker drivers per ear, integrated into the headstrap
• Microphone
  • Dual microphone array

Size and Weight

• Size
  • 175mm x 95mm x 110mm (core module + facial interface)
• Weight
  • 440 g – core module + headstrap
  • 185 g – core module

Software

• Operating System
  • SteamOS 3 (Arch-based)
• Desktop
  • KDE Plasma

Here are the controller tech specs, which come with the headset:

Controller Tech Specs

Controls and Input

• Motion controls
  • Full 6-DOF tracking and IMU support
• Gamepad controls
  • A, B, X, Y buttons (right controller)
  • D-pad (left controller)
  • Full-size magnetic thumbsticks (TMR) with capacitive touch
  • L and R analog triggers
  • L and R bumpers
  • View / Menu / Steam buttons
  • Dual-stage grip buttons
• Haptics
  • Haptic motor in each controller
• Finger tracking
  • Capacitive sensing for all input surfaces
  • Capacitive finger tracking

General

• Connectivity
  • 2.4 GHz link to dedicated headset radio
• Size
  • 126mm x 73mm x 87mm per controller
• Weight
  • With battery: 130 g per controller
  • Without battery: 107 g per controller
• Power
  • One replaceable AA battery per controller
  • 40-hr battery life

At the time of writing, the release date of the upcoming hardware is not known. Valve stated that it is trying to begin shipping them in the beginning of 2026.

Nvidia has officially become the first-ever company to surpass $5 trillion in market value. The company’s inflated market value was pushed higher by the increasing demand for AI, which requires a steady increase in demand for GPU hardware, which Nvidia supplies. Nvidia surpassed $4 trillion in market value only a couple of months ago, with the gaming segment having hit record revenue in the second quarter of 2026. 

Memory makers are prioritizing HBM and server DRAM, leaving less memory for PCs and smartphones, tightening the supply and pushing up prices for retail value. Nvidia recently lined up three major deals, investing up to $2 billion into xAI as a part of a $20 billion agreement for Nvidia’s own AI GPUs, a $5 billion investment in Intel to align NVLink-connected Intel x86 CPUs with Nvidia’s platform, and a plan to invest $100 billion in Open AI while deploying at least 10 GW of AI accelerators, with the first gigawatt expected in H2 of 2026. 

These deals push manufacturing capacity toward AI and reinforce the ongoing DRAM supercycle, which analysts predict to continue at least until 2027. Memory makers hike DRAM prices, which are already pressuring consumer RAM prices. Median DDR5 RAM prices in online retail rose about 15% in the first 10 days of October and are now up 25 to 35% month to date, with more popular kits climbing as much as 50% in less than a month.

Xiaomi says surging memory costs are already pushing up smartphone prices. Micron warns there will be no relief in 2026 as memory supply may continue to trail demand. Adata reports AI data centers are outbidding the PC market, which thins DRAM and NAND inventories and, in turn, raises the risk of RAM and SSD shortages. 

Xiaomi warns that memory inflation is far beyond expectations and could intensify as AI demand tightens hardware component supply. The DRAM supercycle is already lifting PC DDR5 RAM retail prices and is likely to spill over into consumer and gaming devices soon.

In a Weibo post, Xiaomi President Lu Weibing said rising memory costs have reached retail pricing. After consumer pushback, Xiaomi will temporarily reduce the price of the 12/512 GB K90.

The company links the pressure to global AI demand, which has diverted supply and driven up the prices of DRAM and NAND, used in all types of memory and storage, respectively.

Rising prices of gaming and PC hardware

Xiaomi’s warning about phones is an early signal of broader pricing changes that have trickled down from base components to end products.

For gamers, the AI-driven memory squeeze won’t stop at PCs: higher DRAM and NAND costs are likely to push up prices (or trim base configs) on smartphones and handheld gaming PCs next, since these devices rely on the same component supply that’s being tightened by AI build-outs.

On PC RAM specifically, our analysis shows retail DDR5 median prices surged about 15% on October 10 and 30% on October 20 for 2×16 GB and 2×32 GB kits, compared to September. Starting September 22, every week has shown a median increase of 4-6% (and now even more), evidence that upstream DRAM hikes are already reaching consumers.

It is also logical to assume that the rise in DRAM prices will eventually affect GPUs. Graphics cards use GDDR6, GDDR6X, and GDDR7, which share the same supply and manufacturing capacity. The impact usually arrives with a lag, first as higher AIB list prices on mid-range models, then as adjusted MSRPs for new GPUs as OEM contracts expire.

Why RAM is getting more expensive

AI projects are absorbing DRAM budgets and HBM capacity, so suppliers are redirecting wafers, packaging, and logistics toward servers. That tightens supply for conventional DRAM and NAND, lifts contract quotes first, and retail prices follow later.

According to recent reports, major memory manufacturers increased DRAM prices by up to 30% in Q4 2025.

Recent examples of capital and supply commitments that drain memory capacity:

• Nvidia invests $2 billion in xAI, tied to an estimated $20 billion GPU supply arrangement.
• Nvidia commits about $5 billion to Intel CPUs.
• Nvidia may invest up to $100 billion in OpenAI and bring 10 GW of AI GPUs.
• OpenAI signs a new GPU supply deal with AMD.
• Broadcom and OpenAI for 10 GW of custom AI accelerators.

Several analysts now frame this as a DRAM supercycle through 2027: a multi-year period in which demand outpaces supply, driving repeated price hikes. It is not a single hike before the seasonal discounts.

Micron warns there will be no relief in 2026 as supply continues to trail demand across the memory market, implying further RAM price increases.

Adata chairman Simon Chen says AI data centers are outbidding the PC market, cutting upstream DRAM/NAND inventories to roughly 2–3 weeks and raising the risk of RAM and SSD shortages.

DRAM supercycle

In 2017-2019, the DRAM market ran through a similar supercycle: smartphones added more memory per device while cloud providers built out data centers, suppliers kept capex tight, and prices climbed quarter after quarter. Retail DDR4 RAM kits doubled prices in a year. OEMs trimmed RAM in mid-range laptops to hold bills of materials, and consumers either paid more for the same capacity or accepted smaller configs.

By late 2018, the cycle turned: channel inventories built, contract talks slowed, and retail began to ease, with a broader price correction through 2019.

We are seeing the same pattern as 2017-2019: lean inventories and capacity being steered to higher-margin lines (now HBM for AI) leave other memory types exposed to sustained price pressure.

Retail DDR5 RAM prices are already rising, and now OEMs are being hit with hikes too to keep bills of materials in check. Then the market shifted from DDR3 to DDR4; now it is moving from DDR4 to DDR5. Result—fewer discounts and a steady rise in retail prices for mainstream RAM kits.

Micron warns the DRAM market will remain extremely tight with the supply-demand imbalance only worsening through 2026. For consumers, this will result in multiple surges in PC RAM prices and possibly worsen overall memory availability; the first hikes have already reached retail.

According to a DigiTimes report, Micron, one of the largest memory manufacturers, signals that 2026 will not bring any relief for the ongoing memory crisis.

DRAM supercycle will continue, supply will stay tight, the whole memory market situation will become worse, and the supply-demand balance will not reset on its own. This is a memory industry outlook, not a Micron-specific issue.

Emerging memory crisis

A few days earlier, Adata chairman Simon Chen warned that AI data centers are outbidding the PC market and draining DRAM and NAND, leaving only 2-3 weeks of supply and raising the risk of consumer RAM and SSD shortages into 2026.

AI buildouts consume large amounts of memory: GPUs/AI accelerators ship with HBM (stacked DRAM), while the surrounding servers still need substantial quantities of conventional memory.

With thin inventories and demand outpacing capacity, suppliers ration output, raise contract prices, and prioritize higher-margin parts/deals. Because of this, analysts are already predicting a DRAM supercycle through 2027.

DDR5 RAM prices are surging

Retail DDR5 RAM prices have already moved due to hikes from major memory manufacturers.

According to our research, US medians for 2×16 GB and 2×32 GB kits rose roughly 15% by Oct. 10 and around 20-30% by Oct. 20 (compared to September 2025). Weekly prices for those capacities are climbing ~5% for four straight weeks.

AI hardware boom

All this aligns with the extreme pace of AI infrastructure expansion:

• Nvidia to invest up to $100 billion into OpenAI and deploy 10 GW of AI accelerators.
• Nvidia invested $5 billion in Intel, marketed as a buy-in collaboration to develop custom AI chips.
• Nvidia to invest $2 billion in xAI to lock a deal for its own AI GPUs worth $20 billion.
• OpenAI signed a GPU deal with AMD.
• OpenAI lines up a deal to get up to 10 GW chips from Broadcom.

Previous DRAM supercycle

This is not the first time the DRAM market looked like this. In 2017, demand for smartphone and server memory ran ahead of supply for multiple quarters.

Both DRAM spot and RAM retail prices climbed for 1.5 years until inventories were rebuilt. The previous DRAM supercycle sets the pattern for today’s market dynamics.

AI data centers are now outbidding the PC market for core hardware components. Adata chairman Simon Chen says demand is draining DRAM for HBM, NAND for SSDs, and even HDDs, cutting upstream inventories to roughly two to three weeks. As reported earlier, DRAM upcycle is projected through 2027, while consumer RAM prices are already on the rise.

Adata’s CEO, Simon Chen, says the competition for hardware parts has shifted from module makers to cloud and AI buyers, creating simultaneous shortages across DRAM, NAND, and even HDDs that he has never seen before.

Upstream stock has fallen to roughly two to three weeks (versus a more typical two to three months), and Simon has told its sales teams to “sell sparingly” and prioritize long-term customers. HDD scarcity may also nudge some consumers toward SSDs, tightening supply further.

Chen expects tight supply and higher contract pricing to persist into 2026, with DDR4’s production wind-down compounding pressure while capacity is steered toward server demand. Foundry allocation now prioritizes AI servers, followed by general servers, with everything else at the back. This situation is increasing the risk of empty retail shelves and higher prices for consumer RAM and SSDs over the next several quarters.

Rising hardware prices

As noted earlier and reported in our other news articles, hardware prices are rising as AI demand absorbs inventories of core hardware components, while major memory manufacturers are hiking prices. At the same time, analysts predicted a DRAM supercycle at least until 2027. We continue to cover the biggest deals in AI hardware:

• Broadcom and OpenAI deal.
• Nvidia and xAI investment deal.
• Nvidia and Intel investment.
• AMD and OpenAI deal.
• Nvidia and OpenAI investment deal.

Our exclusive analysis shows DDR5 RAM kits’ monthly median price is up to 15% in October (compared to September) at the time of study (October 10) and 20% as of October 15. Over the single last week, 2×32GB DDR5 RAM kits rose 9%. We’re currently expanding the tracking to other PC hardware and will publish weekly updates.

Adata Technology is a Taiwan-based memory and storage manufacturer founded in 2001 by Simon Chen, who still serves as its chairman. The company sells DRAM modules and SSDs for consumer, industrial, and gaming markets under the Adata and XPG brands.

OpenAI and Broadcom will co-develop custom AI accelerators, with plans for approximately 10 gigawatts of deployments, which will start in the H2 2026. Sam Altman’s company also has a 10GW Nvidia partnership; the new deal is meant to lower its reliance on Nvidia’s GPUs over time.

OpenAI signed a multi-year deal with Broadcom to co-develop and deploy roughly 10 gigawatts of in-house AI accelerators and rack systems. OpenAI will design the chips and other systems, while Broadcom will handle development and rollout, with first deployments planned for the second half of 2026 and full buildout by 2029.

This is a long-term capacity plan meant to give OpenAI more control over cost and supply. Separately, the company already has a partnership with Nvidia for the deployment of at least 10 gigawatts of systems, plus $100 billion investment.

The Broadcom deal doesn’t change the near-term picture: OpenAI still relies on Nvidia to run and scale its business through 2026. OpenAI would remain dependent on Nvidia even as its own chip effort progresses, and recent reports suggest OpenAI may lease Nvidia GPUs to reach its 10GW target.

Recent major AI hardware deals

• Nvidia is set to invest $2 billion in xAI, which may then acquire $20 billion worth of Nvidia’s chips.
• Nvidia invests $5 billion in Intel to co-develop SKUs.
• OpenAI signed a deal with AMD for additional GPU supply.
• Nvidia intends to invest up to $100 billion in OpenAI and deploy up to 10 gigawatts of AI accelerators.

These announcements are about AI infrastructure, not consumer graphics. We cover them because they put pressure on base hardware components (DRAM and NAND) and the priorities of memory makers. Leading manufacturers are already hiking prices, driven by demand and potential shortages.

Analysts predicted a DRAM supercycle at least until 2027, with inventories near 2018 lows and vendors steering output toward AI parts. Contract prices are increasing quarter over quarter, and this shift is now impacting retail. Retail RAM and high-capacity SSDs will be repriced first. Other PC parts, consoles, laptops, and handhelds will follow as OEM contracts renew.

DDR5 RAM kit prices have risen about 15% this month, following three consecutive weeks of roughly 5% gains, with the latest week showing even bigger increase. Because gaming GPUs run on DRAM-based GDDR, they’re likely to reprice soon when OEM contracts expire. We’ve completed our consumer RAM price analysis; next, we’re tracking SSDs and GPUs with weekly updates and news posts.

OpenAI is the company behind ChatGPT and the GPT family of models. It trains large AI systems and serves them to consumers and developers, which makes compute supply and cost a core part of its business.

Broadcom is a global semiconductor and infrastructure company: it builds networking chips, storage controllers, and does custom chip manufacturing.

Median DDR5 RAM retail prices have spiked 15% month to date for 2x16GB and 2x32GB kits, and the trend is still upward. For a 2×16GB DDR5 kit, the median price was $95.20 on September 1, $105.60 on October 1, and $132.10 on October 8.

Memory price hikes tied to AI demand have reached consumers, with broad pass-through across all capacities and clock speeds, as the new DRAM supercycle gains momentum.

Disclaimer: methodology is detailed a bit more at the bottom. In short, we analyzed almost a year of daily raw price records for 250+ distinct PC DDR5 kits from Amazon and other major US retailers.

MoM change in DDR5 RAM retail median prices in the US, 2025

October is only 10 days in, but the direction is clear: retail DDR5 prices are rising across all capacities. Weekly data confirm the trend, with DDR5 retail prices of 2x16GB and 2x32GB kits increasing by 4% to 5.5% each week over the past three weeks (see the table below); these 2 groups are the most relevant for gamers, in my opinion.

DDR5 kits are grouped by capacity only. A detailed pass over the data showed that within a single capacity, different clock speeds change the medians only marginally, so to avoid multiplying categories and to keep the charts readable, we present the split by capacity and omit a separate clock breakdown.

2x48GB and 2×64GB kits target beyond the usual consumer/gaming use case; their retail medians track closely, and sample sizes are limited compared to lower capacity RAM, so we put them together as “2×48GB+” for better readability.

Key points:

• Consumer DDR5 prices have risen across all models, with the largest monthly step of about 15% on 2x16GB and 2×32GB RAM kits.
• On slower prices for 2×8GB kits, it’s likely that older inventory is still being cleared, so new pricing filters through gradually.
• The first apparent price move began in late September.
• It appears that October Prime Day and other promotions did not significantly slow price hikes.
• The April anomaly most likely reflects US tariff news.
• October values are month-to-date and will be updated afterward.
• For clarity, the full table is listed below.

Daily median DDR5 RAM retail prices in US, 2025

Despite trimming, 2×32GB remains very volatile due to the highest promo activity. October 9-10 are excluded for now.

No separate lines are provided for 2×48GB and 2×64GB kits because, as said earlier, the samples are smaller compared to other models. We also do not plot a merged “2×48GB+” because capacities fall into different price bands, despite a similar trend, and a combined daily line would be too volatile and make no sense.

Key points:

• Once again, the October spike is clearly visible, and the April spike lines up with tariff news.
• On January 1, a DDR5 RAM 2×8GB kit cost $67, a 2×16GB kit $90.20, and a 2×32GB kit $121.60.
• On July 1, a 2×8GB kit cost $67.30, a 2×16GB kit $88.50, and a 2×32GB kit $146.60.
• Lowest daily median price for 2×8GB kit: $53.50 on February 20.
• Highest daily median price for 2×8GB kit: $83.50 on October 8.
• Lowest daily median price for 2×16GB kit: $78 on February 16.
• Highest daily median price for 2×16GB kit: $132.10 on October 8.
• Lowest daily median price for 2×32GB kit: $99.70 on March 18.
• Highest daily median price for 2×32GB kit: $196.50 on October 8.
• The whole table is too long to display here, sorry.

Why are the prices for PC memory rapidly rising?

The main driver for rising memory costs is AI. We track new AI hardware deals, which occur almost daily at this rate, and those deals absorb DRAM budgets and HBM capacity, which tightens supply for the whole market.

• Nvidia to invest $2B in xAI to lock a $20B deal for its own GPUs.
• Nvidia moved to invest $5 billion in Intel CPUs.
• OpenAI signed a fresh GPU supply deal with AMD.
• Nvidia intends to invest up to $100 billion in OpenAI.

Suppliers prioritize servers and HBM, contract quotes rise, and retail follows. Several analysts now call a DRAM supercycle through 2027. A supercycle refers to a situation where demand consistently exceeds supply over an extended period, causing constant price hikes.

Also, we did not include separate charts or tables, but a smaller European (Germany and the Netherlands) cut of data shows the same DDR5 RAM prices picture with a giant October spike.

Rising hardware prices: What’s next?

The last memory supercycle in 2017/18 ended with consumer RAM roughly doubling in price within a year; we will cover this story in a separate article soon, so you can better understand what’s to come.

And, if you plan a PC upgrade, waiting likely means paying more. The increases will not stop at components: as OEM contracts expire, consumer devices that use memory will move up in price or ship with less capacity (spoiler: they all use memory). There is little good news in the near term.

In the next article, we will look at storage prices.

Methodology

The primary source of raw data is the Keepa API. The dataset covers 250+ distinct DDR5 kits, excluding second-hand and merging duplicates, without preference for specific manufacturers. Data cannot be reversed to individual ASIN listings.

Daily prices were calculated using a rolling median. We used the median instead of the average because, even after removing extreme highs and lows, the mean result remained too volatile. Using the median also lets us combine clock bins and simplify processing, while the result better reflects shelf prices than the average method. The sample size keeps the results representative, and the charts are clearer.

MoM compares the month of October to date (through October 10) with full September medians, and will be updated later. Final values will be higher.

Nvidia is backing xAI as Elon Musk’s company lines up about $20 billion to secure Nvidia AI GPUs for its Colossus 2 data center in Memphis. Nvidia’s own contribution may be up to $2 billion. Confirmed by Nvidia CEO, this deal extends the AI-first trend that keeps hardware prices elevated.

Bloomberg reports that xAI’s financing is structured around an SPV that will purchase Nvidia chips and rent them back to xAI for Colossus 2, the company’s largest site in Memphis.

Sources indicate that the equity slice is approximately $7.5 billion to $8 billion, while the debt portion could reach $12 billion to $12.5 billion. Nvidia is participating on the equity side with up to $2 billion. It’s another significant AI chip deal that secures GPU supply for xAI’s Memphis build, and Nvidia’s shares rose on the news.

Nvidia’s CEO, Jensen Huang, has now confirmed this in a fresh interview, telling CNBC he’s “super excited about the financing”. The other of his quotes is, “The only regret I have about xAI — we’re an investor already — is that I didn’t give him more money.”

Other major recent deals on AI hardware

• Qualcomm said it would acquire Arduino.
• Nvidia moved to invest $5 billion in Intel CPUs.
• OpenAI signed a fresh GPU supply deal with AMD.
• Nvidia intends to invest up to $100 billion in OpenAI.

This is hardware and financing for enterprise data centers and AI-related industries, not consumer graphics, so why do we cover these deals every day? It reinforces the current trend: AI buyers remain at the forefront for advanced packaging and HBM, which keeps pressure on costs and availability across the entire stack, including all kinds of gaming and consumer hardware.

Put simply: for suppliers, AI comes first. If capacity is short and AI costs rise, the consumer hardware market sees higher prices and fewer options as a result.

Memory is already in a DRAM supercycle expected to run through 2027, with inventories near 2018 lows and suppliers prioritizing AI. That pushes contract prices up quarter over quarter, and the increase is reflected in consumer hardware.

Retail RAM and high-capacity SSDs are repriced first, followed by consoles, laptops, and handhelds, which adjust as OEM contracts roll over. Expect fewer promotions, higher list prices, or smaller default capacities.

Qualcomm announced that it will acquire Arduino, and alongside the deal, unveiled the Arduino UNO Q. This $44 board computer pairs a Qualcomm CPU with a real-time microcontroller. Arduino states that the brand, open-source approach, and multi-vendor chip support will remain unchanged.

On October 7, 2025, Qualcomm announced that it had reached an agreement to acquire Arduino, the Italian open-source hardware platform used by a community of over 33 million active users, which “will gain access to Qualcomm Technologies’ powerful technology stack and global reach”.

However, financial terms were not disclosed. Both companies state Arduino will keep its name, tools, and mission, and will continue to support chips from multiple suppliers. The transaction is pending regulatory approval.

Arduino UNO Q

Alongside the announcement, Arduino introduced the UNO Q. The board combines a quad-core Cortex-A53 Qualcomm Dragonwing QRB2210 CPU running Debian Linux with a real-time STM32U585 microcontroller. The base model features 2 GB LPDDR4 and 16 GB eMMC, as well as Wi-Fi 5 and Bluetooth 5.1.

Additionally, The PCB ships with the new Arduino App Lab environment, which enables developers to combine Python apps, Arduino sketches, and lightweight AI models. The price starts at $44, with shipping available from October 24.

In short, Arduino gains Qualcomm’s resource and ecosystem reach, while Qualcomm gets a direct path into a massive DIY user base. However, the reaction of the Arduino user base is mixed for now, primarily due to the platform’s open-source nature, and partly due to the AI angle of the deal.

Qualcomm designs Snapdragon processors and platforms for smartphones, PCs, VR headsets, and other devices. For gaming, Qualcomm offers the Snapdragon Elite Gaming feature set and purpose-built chips like the Snapdragon G3x, which are used in various Android handhelds (Razer Edge, Logitech G Cloud).

Moreover, Meta’s Quest 3 runs on Qualcomm’s XR2 Gen 2 platform. Tying Arduino’s creator base to Qualcomm’s hardware strengthens its reach from mobile and VR into hobbyist controllers, DIY handheld devices, and, as seen in the announcement, cutting-edge AI projects.

Arduino is an open-source hardware and software company/platform for rapid prototyping, used in education, hobby projects, and industry, with a vast, active, and versatile community.

For gaming, this ecosystem underpins a significant share of DIY controllers and other peripherals, as many boards can present themselves as standard USB HID devices (gamepads and keyboards) and are supported by extensive community guides and libraries.

OpenAI and AMD announced a multi-year partnership to deploy up to 6 GW of AMD GPUs for AI infrastructure. The first 1 gigawatt will use MI450 chips and is slated for 2026. The agreement includes a warrant for about 10% of AMD shares, tied to deployment and share-price milestones.

AMD and OpenAI have signed a multi-year, multi-generation deal that begins with the Instinct MI450 and extends to future Instinct products. The first 1 gigawatt build is planned for the second half of 2026; the locations have not yet been disclosed.

The agreement includes a warrant that gives OpenAI the option to buy up to 160 million AMD shares, roughly 10% of the company, if deployment and share-price milestones are met. AMD estimates the total revenue impact over the life of the deal to be tens of billions of dollars.

AMD CEO Lisa Su said: “This partnership brings the best of AMD and OpenAI together to create a true win-win”. OpenAI CEO Sam Altman stated: “AMD’s leadership in high-performance chips will enable us to accelerate progress”.

September AI hardware deals recap

In September 2025, Nvidia inked two framing deals for OpenAI’s compute plans. On September 18, Nvidia announced that it will invest $5 billion in Intel and co-develop multiple generations of data-center and PC products. Intel will design custom x86 CPUs for Nvidia’s AI platforms, provide advanced packaging, and link the chips via NVLink.

On September 22, Nvidia and OpenAI signed a letter of intent to deploy at least 10 GW of Nvidia systems starting in H2 2026, with Nvidia intending to invest up to $100 billion in OpenAI as each gigawatt is deployed. Together, these moves tighten Nvidia’s role in OpenAI’s stack while securing CPU and packaging pathways beyond TSMC.

Put simply: OpenAI secured an option to buy up to 160M AMD shares—days after Nvidia outlined plans to invest up to $100B in OpenAI. I can’t quite put my finger on it.

Nvidia remains the dominant supplier of AI accelerators. The AMD-OpenAI pact diversifies OpenAI’s compute sources and opens a larger AI revenue stream for AMD. The warrant links OpenAI’s potential equity to AMD delivery milestones and share-price thresholds, with reported triggers up to $600 per share.

In the short term, the AMD-OpenAI deal means that both high-bandwidth memory and the newest chip capacity will be allocated to data centers instead of gaming graphics cards. That may keep prices higher for all AMD GPUs, and make top-end ones scarcer.

A broader DRAM upcycle is also pressuring prices. GPUs utilize DRAM as a fundamental component, so while suppliers are steering capacity toward HBM and server memory, contract DRAM and GDDR tend to move up after a short lag. For gaming, this manifests as higher bill-of-materials costs on premium GPUs, tighter VRAM configurations, fewer discounts, and gradually rising prices for new PC builds.