Daily Tech & GNSS News Digest - July 10, 2026

Friday’s tech news put a price tag on the question the market has been arguing about all week: how is the AI buildout actually being paid for? Bloomberg’s answer—$350 billion in debt across the five biggest data-center builders, increasingly raised in euros—landed the same day SK Hynix’s record listing popped 13% and Meta confirmed its in-house chip goes into production in September. And in a small but pointed data point for software investors, Starbucks said it’s using AI to write replacements for the enterprise tools it currently rents. On the positioning side, both stories were about trust infrastructure: Xona started certifying receivers for its LEO signals, and a startup shipped a timing box built on the assumption that GPS will sometimes be gone.

Tech News

Big Tech Has Doubled Its Debt to $350 Billion to Pay for the AI Buildout

Alphabet, Amazon, Meta, Microsoft, and Oracle—the five biggest spenders on new US data centers—have collectively added some $350 billion to their debt obligations over the last five years, doubling their debt load to finance the AI spending spree, Bloomberg reports. The five paid a combined $10 billion in debt interest last year, more than double 2019’s bill—still small next to cash flows like Google’s roughly $64 billion last quarter, but growing fast enough to reshape credit markets.

The more striking detail is where the money is coming from: American tech has effectively run short of dollars to borrow, so the hyperscalers have turned to Europe, with Morgan Stanley expecting their euro-denominated borrowing to reach €50 billion this year, per The Next Web’s read of the data. Days after Amazon’s $25 billion bond sale forced investors to dump existing tech debt to make room, the aggregate picture confirms what that sale hinted at: the AI buildout’s financing needs have outgrown any single market.

Update: SK Hynix Pops 13% in Its Nasdaq Debut—and Immediately Spends $8.6 Billion of It at ASML

SK Hynix closed its first day of US trading up 13% at $168.01 after opening at $170, capping the $149-per-ADS offering covered here yesterday with a $26.5 billion raise—the largest-ever US listing by a foreign company and, per Money Morning, the second-largest US listing in history behind SpaceX’s debut four weeks ago. Chairman Chey Tae-won told CNBC that “demand is enormous”; the ADRs trade under a temporary ticker before switching to SKHY next week.

The proceeds are already moving: SK Hynix committed $8.6 billion to acquire advanced EUV lithography equipment from ASML, per reporting on the debut—converting Wall Street’s AI-memory enthusiasm directly into the machines that make high-bandwidth memory. Bloomberg frames the listing as a pure-play bet on the AI boom, and the first session suggests US investors see it the same way.

Meta’s In-House “Iris” Chip Enters Production in September as Compute Doubles to 14 Gigawatts

Meta will put its custom AI chip, code-named Iris, into production at TSMC starting in September, according to an internal memo reviewed by Reuters. The chip—part of the fourth generation of Meta’s MTIA accelerator program, with Broadcom assisting on design—cleared testing in just six weeks with no major issues found, and anchors a plan to double Meta’s computing infrastructure from 7 gigawatts this year to 14 gigawatts in 2027, backed by a raised 2026 capex forecast of $125–145 billion.

The supply chain tells its own story: TSMC fabrication, Samsung RAM, SanDisk storage, and Sumitomo Electric fiber, per Reuters. Two days after Meta shipped its first paid model priced to undercut OpenAI and Anthropic, the chip timeline shows the other half of that strategy—owning enough of the inference stack to make aggressive pricing sustainable. Meta shares headed for their best week since early 2024 on the news, per CNBC.

Starbucks Is Using AI to Write Its Own Software—and Software Vendors Noticed

Starbucks is developing in-house, AI-built systems to replace software it currently buys from Big Tech vendors—including a Microsoft inventory-monitoring system, an IBM maintenance-management tool, and Oracle’s Simphony point-of-sale platform, Bloomberg reports. CTO Anand Varadarajan told employees the chain spends roughly $400 million a year on software, and the replacement push is part of a $2 billion cost-reduction target in its turnaround plan, with some homegrown tools rolling out by the end of next year pending testing.

Software stocks slid on the report, per Yahoo Finance—because the fear isn’t about one coffee chain. If generative AI lets ordinary enterprises build tailored replacements for rented enterprise software, the recurring-revenue base underpinning the entire sector gets renegotiated. Coming the same week the buy-chips-sell-software trade started wobbling, it’s a concrete example of the thesis software bulls have been dreading.

Additional Headlines:

• The EU rules Meta’s designs are “addictive”: Regulators found Instagram and Facebook in breach of EU law over engagement-maximizing design choices, opening another regulatory front for Meta in an otherwise strong week, per CNBC’s TechCheck roundup.
• Circle gets the green light to operate as a bank: The stablecoin issuer’s approval marks one of the clearest steps yet in crypto’s absorption into the regulated US financial system, per CNBC.
• Beta Technologies wraps first flights in the US government’s air-taxi pilot: The electric-aviation company completed its initial test flights under the federal eVTOL pilot program, per CNBC.
• xAI launches Grok 4.5: The new model touts a 4.2x token-efficiency gain and aggressive $2 per million token pricing, joining a week in which Meta and OpenAI both moved on the price-performance frontier, per BuildFastWithAI’s roundup.

GNSS News

Xona Starts Certifying the Pulsar Ecosystem—With Trimble, Septentrio, and STMicroelectronics Signed On

Xona Space Systems launched Pulsar Verified, a certification program that validates receivers, chipsets, and test equipment for interoperability with its Pulsar LEO PNT signals, per Inside GNSS and SpaceNews. The inaugural cohort is a who’s-who of the receiver industry: Trimble (bringing devices launched as early as 2018 to Pulsar compatibility), Septentrio (part of Hexagon, with its LeoRx receiver progressing through verification), STMicroelectronics (which has demonstrated a Pulsar receiver on its TeseoV chipset), plus Safran, StarNav, and Keysight.

Xona pitches the program as a response to escalating jamming and spoofing hitting aviation, shipping, agriculture, and financial timing—claiming its stronger LEO signal can shrink a jammer’s effective area by as much as 95% in live-sky tests. The certification push follows the company’s $170 million Series C in March and a new Burlingame, California satellite integration facility opened in April, as Xona scales toward a planned 258-satellite constellation. A signal is only as useful as the receivers that trust it; this is Xona building that trust layer before the full constellation flies.

Jericho Receivers Ships a Software-Defined Atomic Clock That Doesn’t Need GPS to Stay Honest

Jericho Receivers, a subsidiary of All 6G, commercially launched its Software-Defined Atomic Clock (SDAC), which fuses GNSS timing with terrestrial ATSC 3.0 broadcast signals—including the Broadcast Positioning System (BPS)—and network-based precision time protocols, per Inside GNSS and Business Wire. Rather than relying on conventional holdover, where a device drifts on its internal clock during a GNSS outage, the SDAC continuously cross-references all three independent time sources, maintaining atomic-clock-grade synchronization through jamming, spoofing, or outright signal loss.

The approach is notable for what it leans on: America’s next-generation TV broadcast standard as a GPS-independent timing backbone. With the FCC advancing BPS as a complement to GPS and critical-infrastructure operators under pressure to harden their timing, fusing broadcast towers into the reference mix offers resilience without waiting for new satellites. The SDAC is available now for evaluation and deployment—another entrant in the fast-crowding market for timing that survives a bad GPS day.

---

Key Takeaways

• The AI buildout is now a credit-market story: The five biggest data-center builders have doubled their debt to $350 billion and are raising €50 billion a year in Europe as dollar markets strain—while SK Hynix’s 13% debut pop shows equity investors still funding the trade enthusiastically.
• Vertical integration is accelerating at both ends of the stack: Meta’s Iris chip hits TSMC production in September to double compute to 14 gigawatts, while Starbucks uses AI to replace Microsoft, IBM, and Oracle software—chips and enterprise apps both being pulled in-house.
• PNT resilience is becoming an ecosystem play: Xona’s Pulsar Verified program enlists Trimble, Septentrio, and STMicroelectronics to certify receivers before its LEO constellation fully flies, and Jericho’s software-defined atomic clock fuses GNSS, TV broadcast, and network timing so infrastructure keeps ticking when GPS doesn’t.