Daily Tech & GNSS News Digest - June 24, 2026

Yesterday the chip trade cracked; today it found its footing again—and the day’s news read like a referendum on who controls the memory, the silicon, and the supply chain underneath the AI boom. SK Hynix is bidding to become the boom’s next public-market giant, OpenAI is building its own chip to stop renting all of Nvidia’s, and Micron answered the selloff with numbers too big to argue with. On the positioning side, the throughline was sovereignty again: a GNSS chip built end-to-end inside Europe, and a regulator spelling out for pilots exactly how to tell when their satellites have gone dark.

Tech News

SK Hynix Files for a $29.4 Billion US Listing—Poised to Be the Second-Largest IPO Ever

SK Hynix, the South Korean memory maker that has emerged as the top supplier of high-bandwidth memory (HBM) for AI data centers, filed for a US listing aiming to raise up to $29.4 billion (45.45 trillion won), Bloomberg reported June 24. If it hits that target, the offering would surpass Alibaba’s $21.8 billion debut in 2014 to become the second-largest US IPO in history. The company plans to sell up to 17.79 million shares and expects to begin trading on the Nasdaq around July 10, with Bank of America, Citigroup, Goldman Sachs, and JPMorgan leading the deal, per SiliconANGLE and Renaissance Capital.

The listing is a victory lap for a come-from-behind story—SK Hynix’s stock has climbed roughly 300% as HBM became the single most supply-constrained component in the AI buildout. Proceeds will fund manufacturing capacity, including four new fabs at the Yongin Cluster rising near Seoul. Coming days after SpaceX’s own capital maneuvers, the filing underlines how the scramble to finance AI infrastructure is now playing out at the scale of record-breaking public offerings.

OpenAI and Broadcom Unveil ‘Jalapeño,’ Their First Custom Inference Chip

OpenAI introduced Jalapeño, its first custom AI processor, co-developed with Broadcom and built specifically to run—rather than train—large language models, the two companies announced June 24. Per OpenAI, Tom’s Hardware, and VentureBeat, the reticle-sized ASIC went from initial design to manufacturing tape-out in just nine months, a pace OpenAI credits partly to using its own models to accelerate the design work. Engineering samples are already running ML workloads in the lab at production frequency and power, including GPT-5.3-Codex-Spark.

The strategic intent is unmistakable: reduce reliance on Nvidia by owning the inference layer outright. OpenAI says Jalapeño delivers performance-per-watt “substantially better” than current state-of-the-art and is the first step in a multi-generation platform, pairing OpenAI-designed accelerators with Broadcom’s silicon, networking, and Celestica’s system integration. Deployment is slated to begin by the end of 2026, scaling toward gigawatt-class data centers with Microsoft and other partners. TechRadar framed it bluntly as an Apple-style bid to “build the full stack”—the latest sign that the biggest AI buyers no longer want to be just customers of the chipmakers.

Update: Micron Roars Back From the ‘Chip-Wreck’ With a Blowout Quarter

One day after memory names led a brutal semiconductor selloff, Micron delivered the rebuttal. The company reported fiscal Q3 revenue of roughly $41.5 billion—quadruple a year earlier—with profit surging from $1.88 billion to $28.2 billion, and guided fourth-quarter revenue to $49–51 billion, far above the ~$43.2 billion analysts expected, per Bloomberg, CNBC, and Yahoo Finance. Shares jumped more than 13% in late trading, recovering the ground lost in the prior session’s rout. The new detail beyond the headline: Micron also declared a dividend and tied its outlook to “insatiable” memory demand from AI.

The results land alongside a strategic agreement with Anthropic unveiled June 22 that Micron didn’t dwell on during the selloff—a four-pillar deal spanning HBM and storage co-design, a multi-year supply agreement, Claude adoption across Micron’s own operations, and a strategic investment in Anthropic’s Series H (the $65 billion round that valued the lab at $965 billion), per Micron’s release and StockTitan. For a stock that cratered ~13% on Tuesday over fears the AI capex cycle was peaking, the message Wednesday was that, for memory at least, demand is still running well ahead of supply.

Additional Headlines:

Google’s $100 Home Speaker brings Gemini into the living room: Google launched its first standalone smart speaker in nearly six years—$99.99, shipping June 25—built around “Gemini for Home” with multi-step conversations, mid-sentence corrections, and 10 new voices, per Google’s blog and Engadget. Bloomberg’s review called the AI impressive but the audio “underwhelming.”
Valor Equity targets a $2.5 billion Fund VII: Elon Musk-aligned Valor Equity Partners is raising at least $2.5 billion for its seventh fund, a portion already earmarked for further SpaceX investment, per Bloomberg via TechCrunch—another channel funneling private capital into the AI-and-space buildout.

GNSS News

GlobalFoundries and Qualinx Tape Out the First Fully European GNSS Chip

GlobalFoundries and Dutch firm Qualinx completed what they describe as the first fully European, end-to-end manufacturing flow for a GNSS system-on-chip, a concrete step toward supply-chain sovereignty for positioning-dependent aerospace, defense, and critical-infrastructure systems, per Inside GNSS and GPS World. The milestone centers on Qualinx’s QLX3xx, a reconfigurable GNSS SoC and analog front end aimed at secure PNT, resilient timing networks, and ultra-low-power receivers for edge devices—designed, taped out, and fabricated entirely at GF’s Dresden fab on its FDX process.

The security claim is the whole point: GF says no design data or physical materials left the EU at any stage of production, and it’s working with Deutsche Telekom to keep even the production data on European networks and cloud infrastructure. GF aims to have a fully automated “trusted European flow” in place by end of 2026, with foundry engagements opening to aerospace and defense customers in 2027. It’s the GNSS-chip analog of the sovereignty anxieties echoing across the rest of the week’s news—when positioning underpins defense and critical infrastructure, who fabricates the silicon becomes a strategic question, not a procurement one.

FAA Hardens Its Jamming-and-Spoofing Playbook for Pilots

The FAA updated its GNSS Interference Resource Guide to Version 1.1, sharpening guidance on how interference degrades flight-deck systems and how pilots can recognize it in the cockpit, per GPS World, Aviation Week, and AIN. Building on the first edition released in December 2025 and incorporating recommendations from an industry-government GPS/GNSS Disruption Action Team, the revision spells out that interference can degrade or disable far more than basic navigation—RNAV/RNP, ADS-B, CPDLC, automatic navaid tuning, synthetic vision, HUDs, and TAWS are all on the list.

Notably, the guide now maps the worst spoofing hotspots, where GPS-like signals are being acquired by civil-aviation receivers in place of the real thing: the Eastern Mediterranean, Black Sea, Russia and the Baltic region, the India/Pakistan border, Iraq and Iran, the Korean peninsula, and areas around Beijing. That a national aviation regulator now publishes a geography of GNSS deception—and detailed cockpit symptoms to match—captures how thoroughly interference has shifted from edge case to standing operational hazard.

Key Takeaways

AI’s capital cycle is now public-market scale: SK Hynix’s bid for a $29.4 billion listing—on pace to be the second-largest US IPO ever—shows the HBM crunch minting a new giant, while Valor’s $2.5 billion fund keeps private money flowing into the same buildout.
The biggest AI buyers want to own the silicon: OpenAI and Broadcom’s Jalapeño inference chip, taped out in nine months, is a direct play to escape Nvidia dependence—and Micron’s blowout quarter and Anthropic tie-up prove memory demand is still outrunning supply a day after the “chip-wreck.”
Sovereignty is the GNSS story too: A fully European GNSS chip with no design data leaving the EU, and an FAA guide now mapping global spoofing hotspots, point the same direction—when positioning underpins defense and aviation, both the fab and the airspace need hardening.