Daily Tech & GNSS News Digest - May 26, 2026

US tech stocks shrugged off weekend escalations in Iran and Ukraine to set fresh record highs, led by a memory-chip rally that pushed Micron past a trillion-dollar valuation and a fresh Qualcomm AI-data-center coup that expanded its ASIC pipeline into China — while in the positioning world, software-defined GNSS and Cold War–era terrestrial signals both moved a step closer to plugging the gaps left by jammed and spoofed satellites.

Tech News

Qualcomm Strikes Multi-Million-Chip AI Deal With ByteDance

Qualcomm has reached an agreement to supply TikTok parent ByteDance with millions of custom AI application-specific integrated circuits for the Chinese giant’s data centers, Bloomberg reported May 26. The deal extends beyond chip sales: Qualcomm will also help shepherd a proprietary ASIC that ByteDance has already designed through the production process, deepening a partnership that gives the San Diego chipmaker a high-volume entry point into one of the fastest-growing segments of the semiconductor industry.

The arrangement is structured to stay within US export rules — Qualcomm’s chips fall under the legally acceptable computing thresholds that have governed AI-chip shipments to Chinese customers since the H200 framework was approved. That makes ByteDance one of the first major customers for Qualcomm’s AI-focused ASIC line, broadening a roster that already includes Saudi Arabia’s Humain as the launch customer for the AI200 data-center processor unveiled at CES 2026.

Investors greeted the news aggressively: Qualcomm shares rose as much as 8.3% to a new intraday record, capping a months-long re-rating as the company has pivoted from a smartphone-modem story into a credible challenger in the inference-chip market long dominated by Nvidia. With Anthropic also exploring Microsoft’s Maia silicon and Google pushing TPU 8, the data-center accelerator market is quickly becoming a multi-vendor fight rather than a one-name trade.

Micron Crosses $1 Trillion as HBM4 Sells Out and UBS Triples Its Target

Micron Technology joined the trillion-dollar club for the first time on May 26, with shares surging as much as 19.3% to an intraday record of $891.27 before settling at $872.24. The move made Micron the best-performing stock in both the S&P 500 and the Nasdaq, and crystallized a stunning twelve-month run in which memory has decoupled from the broader semiconductor cycle on the back of unrelenting AI-server demand.

The trigger was UBS analyst Timothy Arcuri raising his price target from $535 to $1,625 — a 204% increase and the highest among 46 brokerages tracked by LSEG. Arcuri’s note pointed to disclosures from CEO Sanjay Mehrotra that Micron’s entire 2026 HBM4 supply is now sold out under long-term, fixed-price contracts, and that the company is filling only 50–65% of its key customers’ medium-term demand. That is pricing power of a kind the famously cyclical memory industry has rarely if ever experienced.

Memory’s reordering of the AI-chip leaderboard was the broader story of the session: 16 of the top 20 S&P 500 gainers were semiconductor or computer-hardware names, with tech, industrials, and materials sectors driving the index higher even as front-month WTI crude fell 4.7% to $92.94 on softer demand readings. The Russell 2000 crossed 2,900 for the first time on a 1.77% gain, an unusual breadth signal in a rally that has often been narrowly led by mega-cap AI names.

STMicroelectronics Pushes Gallium Nitride Into the AI Power Stack

STMicroelectronics unveiled a new generation of gallium nitride (GaN) power semiconductors on May 26, positioned squarely at the AI-server, EV, and robotics power-conversion bottleneck that has emerged as data-center electricity demand outruns conventional silicon. The new PowerGaN family offers low conduction losses, very low switching loss at high operating frequencies, and zero reverse-recovery charge — characteristics that let designers shrink power-conversion stages while moving more current at higher efficiency.

For hyperscalers, the timing matters: AI racks built around Nvidia Vera Rubin and Google TPU 8 systems are pushing rack-level power into the hundreds of kilowatts, with conversion losses showing up as both a thermal and a sustainability problem. Wide-bandgap chemistries like GaN — and silicon carbide further upstream — are increasingly the path to keeping efficiency curves moving in the right direction. ST’s announcement lands the same week that Micron’s HBM constraints and Nvidia’s supply warnings have made every link in the AI compute chain a focus for investor scrutiny.

Additional Headlines

Markets brush off Iran strikes: The Nasdaq (+1.17%) and S&P 500 (+0.61%) closed at record highs despite weekend US strikes on Iran and renewed Ukraine escalation, with oil sliding rather than spiking as traders bet that supply disruption risk was overstated.
Trump renews 25% Apple tariff threat: The administration revived a threat to impose a 25% tariff on Apple and other device makers unless they shift more manufacturing to the United States, with Apple shares down roughly 20% year-to-date and the company already accelerating an iPhone production shift to India.
Anthropic’s compute bill emerges from SpaceX S-1: Filings around last week’s SpaceX IPO disclosed that Anthropic will pay $1.25 billion a month through May 2029 — up to $45 billion in total — for compute at the Colossus 1 data center near Memphis, with the deal expanding to Colossus 2 as the Claude maker scales training.
OpenAI says model cracked an 80-year-old geometry problem: One of OpenAI’s general-purpose reasoning models autonomously produced a proof to a long-standing open problem in geometry, the company said — early proof points for AI’s “scientific assistant” pitch that investors are pricing into the next funding round.

GNSS News

FocalPoint Positioning Launches Precise+ for Sub-Metre Accuracy Without Sensor Fusion

UK software-defined receiver specialist FocalPoint Positioning unveiled Precise+ at the European Navigation Conference in Vienna on May 26, claiming sub-metre GNSS accuracy in environments that typically defeat conventional carrier-phase tracking — and notably, without inertial sensors, dead reckoning, or sensor fusion. The system attacks the cycle-slip problem that causes RTK and PPP solutions to degrade or drop out when signals are blocked or reflected by buildings, foliage, or terrain.

In field testing at Thetford Forest in dense canopy, FocalPoint says Precise+ delivered 80 cm accuracy at the 99th percentile — performance that RTK in particular struggles to deliver outside open-sky conditions even with high-end inertial backup. The company is built around its patented Supercorrelation signal-processing platform, which extracts more usable information from each tracked satellite rather than relying on additional sensors to bridge outages.

The commercial target is unambiguous: automotive ADAS, automated driving, and V2X, where lane-level positioning is required across the same urban canyons, tunnels, and tree-lined approaches that wreck consumer GNSS. FocalPoint says it is working with chipset manufacturers on commercialization but has not disclosed a launch timeline or pricing. The launch lands alongside u-blox’s recent ZED-X20P Galileo High Accuracy Service push and Septentrio’s new mosaic-G5 P8 — three different industry bets on how to engineer the next layer of precision into receivers without leaning on external networks or sensors.

UK MoD Awards £6 Million eLoran Contract to QinetiQ-Led Team Elaris

The UK Ministry of Defence has awarded a £6 million contract to Team Elaris — a consortium led by QinetiQ with UrsaNav, Roke, and GMV — to develop a deployable concept for enhanced Long-Range Navigation (eLoran) as a GPS-independent positioning, navigation, and timing alternative for military operations in contested environments, Inside GNSS reports. The two-year programme runs under the MoD’s Urgent Compass initiative and focuses on rapidly deployable terrestrial PNT for worldwide use.

eLoran is the modernized descendant of the Cold War-era Loran-C system. Because it relies on high-power, low-frequency terrestrial signals rather than weak satellite broadcasts from medium Earth orbit, it is significantly harder to jam or spoof than GNSS, and it remains usable when satellite coverage is denied or degraded. The contract complements QinetiQ’s parallel role in the UK’s broader Robust Global Navigation System programme, signalling that British defence planners are pursuing a layered resilience strategy across terrestrial, LEO, and traditional GNSS.

The award fits an unmistakable pattern: with Norway’s DNK paying for Iridium-based maritime PNT, the US Army funding NorthStar non-RF positioning, ESA validating Celeste LEO-PNT, and Northrop Grumman fielding production M-code anti-jam systems, governments are no longer arguing about whether GPS needs a backup. They are placing simultaneous bets across every plausible signal stack — and writing real cheques to find out which ones actually scale.

Key Takeaways

The AI-chip leadership map is being redrawn in real time: Qualcomm landing ByteDance for custom data-center ASICs and Micron crossing a $1 trillion valuation on sold-out HBM4 — with UBS arguing the stock is still less than half of fair value — show that the AI buildout’s biggest beneficiaries in 2026 are increasingly the inference and memory layers, not just GPUs.
The power stack matters as much as the silicon: STMicroelectronics’ new GaN family is a reminder that as rack-level demand climbs into the hundreds of kilowatts, AI economics depend on conversion-efficiency gains that hyperscalers can’t get from incremental silicon improvements alone.
GPS resilience is becoming a layered architecture, not a single product: FocalPoint’s signal-processing approach to sub-metre accuracy without sensor fusion, paired with the UK MoD’s £6 million bet on eLoran terrestrial timing, shows that the post-GPS positioning era is being built simultaneously inside the receiver, on the ground, in the air, and in orbit — with no single solution expected to do the whole job.