Daily Tech & GNSS News Digest - May 27, 2026

The AI buildout kept rewriting the leaderboards this week as two memory chipmakers crossed the trillion-dollar mark and Marvell’s custom-silicon business set fresh records — while Beijing widened its grip on private-sector AI talent. In the positioning world, regulators and space agencies are racing to harden navigation against a growing interference threat, with the FAA refreshing its jamming guidance and ESA’s first dedicated LEO-PNT satellites coming online.

Tech News

SK Hynix and Micron Join the $1 Trillion Club on AI Memory Boom

South Korea’s SK Hynix became the third Asian company ever to surpass $1 trillion in market capitalization on May 27, with shares jumping 9.3% in Seoul to push its 12-month gain past 1,000%. The milestone came one day after Micron Technology crossed the same threshold, after the Boise-based memory maker soared 19% on May 26 — its biggest single-day gain since 2011 — following a UBS note arguing the stock could double again over the next year.

The twin breakouts cement memory as the second pillar of the AI infrastructure trade, alongside Nvidia’s accelerators. Explosive demand for high-bandwidth memory used in AI training racks has created a global shortage that SK Hynix, Samsung, and Micron — the three dominant HBM suppliers — are still struggling to fill. Bloomberg notes that SK Hynix’s run has now made it the third Asian firm in the trillion-dollar club, alongside TSMC and Tencent, while Micron has joined the megacap ranks of Apple, Nvidia, and Alphabet.

The rally also marks a striking shift in how investors value commodity-like memory businesses that were, until recently, treated as cyclical also-rans. With Nvidia warning that GPU supply will be constrained throughout the Vera Rubin cycle and hyperscalers locked into multi-year capex commitments, the memory tier of the AI stack has become a structural bottleneck — and the chipmakers who own it are being repriced accordingly.

Marvell Posts Record Quarter as Custom AI Silicon Scales

Marvell Technology reported fiscal Q1 2027 revenue of $2.418 billion on May 27, a record that grew 28% year over year and beat consensus, with non-GAAP earnings of $0.80 per share against the $0.75 expected. Operating cash flow hit a record $638.8 million, and the company guided Q2 revenue to roughly $2.7 billion at the midpoint — implying 35% annual growth as its custom AI silicon ramps with hyperscale customers including Amazon and Microsoft.

The print also reflects two strategic acquisitions that closed in February: Celestial AI, which brings optical interconnect IP for scaling AI clusters beyond electrical limits, and XConn Technologies, a specialist in PCIe and CXL switching. Together they round out Marvell’s pitch as the go-to merchant supplier for custom accelerators and the networking fabric stitching them together — the layer just below Nvidia in the AI stack that hyperscalers increasingly want to own.

Beijing Extends Travel Curbs to AI Talent at Alibaba and DeepSeek

China is now requiring top AI researchers at private firms — including Alibaba and DeepSeek — to obtain government approval before traveling abroad, Bloomberg reported May 26. The policy was first applied quietly to some DeepSeek executives in December 2025 and has since been broadened to the wider private AI sector, marking a notable escalation in Beijing’s effort to safeguard sensitive technology and personnel.

China has long restricted overseas travel for nuclear scientists, prominent academic researchers, and state-enterprise executives, but extending the regime to private companies signals just how strategic AI work at firms like Alibaba and DeepSeek has become in the eyes of Chinese authorities. The move adds friction to international research collaboration and may complicate global recruiting at a moment when US labs are still aggressively bidding for top Chinese-origin AI talent.

Additional Headlines

JPMorgan to hire more AI staff, fewer bankers: CEO Jamie Dimon told Bloomberg the bank’s workforce will shift toward engineers, data scientists, and AI specialists, with annual attrition of 25,000–30,000 employees giving JPMorgan room to retrain and redeploy rather than execute large layoffs.
Big Tech AI capex on track for $665B in 2026: Alphabet, Amazon, Meta, and Microsoft are collectively guiding to roughly $665 billion of AI-related capital expenditure this year — nearly 75% above 2025’s $381 billion — with Alphabet alone raising 2026 capex to $180–190 billion.
Anthropic plants a flag in Korea: Anthropic appointed KiYoung Choi as Representative Director of Korea on May 26, formalizing its expansion ahead of opening a Seoul office aimed at enterprise customers and AI safety collaboration in one of the world’s most chip-intensive markets.

GNSS News

FAA Refreshes GNSS Interference Resource Guide as Spoofing Hotspots Multiply

The FAA released Version 1.1 of its GPS/GNSS Interference Resource Guide on May 20, the agency’s most substantive update yet to the document operators and pilots use to navigate a worsening jamming and spoofing landscape. The refresh, developed by the FAA’s AFS-400 Flight Technologies and Procedures Division with input from the Performance Based Operations Rulemaking Committee’s GPS/GNSS Disruption Action Team, focuses on current interference trends, system-level impacts on aircraft avionics, and updated pilot procedures and training recommendations.

Notably, the new guide names the top spoofing hotspots that flight crews now have to plan around: the Eastern Mediterranean, the Black Sea, Russia and the Baltic region, the India–Pakistan border, Iraq and Iran, the Korean Peninsula, and areas around Beijing. That list, all but unthinkable five years ago, captures how thoroughly GNSS jamming and spoofing have shifted from niche electronic-warfare concern to a routine factor in commercial flight planning. The update also folds in lessons from the broader maritime interference picture, including the more than 1,100 vessels documented by Windward as experiencing GPS and AIS disruption over a single 24-hour window in the Strait of Hormuz earlier this year.

The new guide arrives as the FAA, EASA, and ICAO converge on more aggressive interference-reporting requirements, and as airlines push for resilient PNT options spanning multi-constellation receivers, inertial navigation, and emerging LEO-based signals. With Starlink’s recent shutdown of its consumer-grade location API removing one informal backup and the Pentagon’s Resilient GPS smallsat layer cancelled earlier this year, regulators are increasingly leaning on procedures and training to bridge the gap until purpose-built resilient PNT hardware reaches scale.

ESA’s First Celeste LEO-PNT Satellites Operational at 510 km

The European Space Agency’s first two Celeste in-orbit demonstration satellites — IOD-1 and IOD-2, launched from Rocket Lab’s Complex 1 in New Zealand — are now operational at roughly 510 kilometers altitude, marking the program’s transition from paper architecture to live signal experimentation. The CubeSats will transmit across UHF, L-band, S-band, and C-band, validating a multi-frequency design intended to deliver positioning resilience that current MEO constellations like GPS and Galileo cannot match on their own.

Pietro Giordano, ESA/ESTEC’s LEO-PNT System Manager, has said the demonstration phase “is fundamental for us to master the technology” ahead of an in-orbit preparation phase planned for 2027–2028 and operational European LEO-PNT services targeted for 2032. Development is led by GMV and Thales Alenia Space France, covering satellites, ground infrastructure, and system specifications, with planned integration into 5G and terrestrial networks pointing toward indoor and urban applications that today’s GNSS signals cannot reliably serve.

Celeste is one of several LEO-PNT efforts now racing to first signal: Xona Space Systems is building out its Pulsar constellation with US commercial customers, and China continues to expand low-orbit positioning experiments. Together they reflect a broader industry conviction that the answer to jamming-saturated MEO bands is not just better receivers, but stronger, lower, and more numerous signals from orbit.

Key Takeaways

Memory is the new center of gravity in AI hardware: SK Hynix and Micron breaching $1 trillion days apart confirms that high-bandwidth memory has joined GPUs as a structural bottleneck — investors are now repricing the whole HBM tier as a strategic chokepoint rather than a commodity.
The AI talent map is fracturing along national lines: Beijing’s extension of travel curbs to private-sector AI researchers at Alibaba and DeepSeek, on top of US export controls and chip restrictions, signals that the next phase of AI competition will be defined as much by personnel mobility as by silicon.
GNSS resilience is now a procedures-and-LEO problem: With the FAA cataloguing seven distinct spoofing hotspots and the Pentagon’s smallsat backup off the table, regulators are leaning harder on pilot training while space agencies like ESA push purpose-built LEO-PNT toward operational service later this decade.