Daily Tech & GNSS News Digest - May 5, 2026

AI spending ambitions reached staggering new heights as OpenAI disclosed a $50 billion compute budget for this year alone — while on the hardware side, Apple quietly explored breaking its TSMC dependency and an AI chipmaker prepared the year’s biggest IPO.

Tech News

OpenAI Plans to Spend $50 Billion on Compute This Year

OpenAI co-founder and President Greg Brockman revealed during court testimony on Monday that the company expects to spend $50 billion on computing power in 2026 to support its AI models. The disclosure came during Brockman’s testimony in the ongoing lawsuit brought against OpenAI, offering a rare window into the financial scale of frontier AI development.

The figure underscores the extraordinary capital intensity of the AI race. To put it in context, OpenAI would need to consume two gigawatts of Amazon’s Trainium AI accelerators and deploy its top GPT models across AWS infrastructure just to account for $35 billion of the spending committed by Amazon’s cloud arm. Bloomberg reports that the $50 billion figure, if representative of total compute spend across cloud and on-premise purchases, implies massive demand for high-performance accelerators, networking, and power infrastructure that will ripple across the semiconductor supply chain for years.

Apple Explores Intel and Samsung for US Chipmaking

Apple has held exploratory discussions with Intel and Samsung Electronics about manufacturing the main processors for its devices in the United States, Bloomberg reported on Monday. Apple executives have visited a Samsung plant under development in Texas and held separate preliminary talks with Intel about enlisting its foundry services — a move that would offer a secondary option beyond longtime manufacturing partner TSMC.

The motivation is twofold: reducing supply chain risk if geopolitical tensions over Taiwan escalate, and strengthening Apple’s relationship with the Trump administration, which has championed Intel as a national semiconductor champion. However, neither effort has resulted in orders, and Apple remains cautious — Intel and Samsung cannot yet reliably match the production scale and yield that TSMC delivers. The talks highlight how even Apple’s $111 billion quarterly revenue machine feels the pressure to diversify away from a single-source dependency.

GameStop Makes $56 Billion Bid for eBay

In one of the most audacious deals in recent memory, GameStop submitted an unsolicited $56 billion offer to acquire eBay — a company nearly four times its size. CEO Ryan Cohen proposed $125 per share, half in cash and half in GameStop stock, representing a 46% premium to eBay’s closing price on February 4, when GameStop began building a 5% stake.

GameStop plans to fund the deal through cash reserves and up to $20 billion in debt financing committed by TD Securities, pledging $2 billion in annual cost savings within 12 months of closing. Cohen said he would serve as CEO of the combined company, positioning the merger as a path to building a “legit competitor” to Amazon. eBay’s board said it had received no prior outreach from GameStop and would review the proposal “with a focus on the value to be delivered to eBay shareholders.” GameStop shares fell on the announcement, reflecting investor skepticism about the deal’s feasibility.

Additional Headlines

Cerebras targets $3.5 billion IPO at $26.5 billion valuation: The AI chipmaker filed to sell 28 million shares at $115–$125 each on Nasdaq, with Bloomberg reporting banks have already fielded $10 billion in orders — nearly 3x oversubscribed before pricing.
Global semiconductor sales hit $298.5 billion in Q1 2026: An increase of 25% compared to Q4 2025, driven by surging AI infrastructure demand and TSMC’s accelerating 2nm production expansion.
Nvidia’s market cap reaches record $5.26 trillion: The chipmaker continues to dominate the AI accelerator market despite growing competition from cloud providers developing custom silicon.

GNSS News

u-blox Begins Integrating LEO Satellite Signals into Its GNSS Platform

u-blox has launched a technical program to evaluate how Low Earth Orbit navigation signals can be fused with traditional GNSS on its X20 all-band positioning platform. The work, conducted under ESA’s Navigation Innovation and Support Program (NAVISP Element 2), follows the successful launch of the first two Celeste LEO-PNT demonstration satellites on March 28, 2026 from New Zealand aboard a Rocket Lab Electron vehicle.

The project focuses on three areas: characterizing emerging LEO signal transmissions across L- and S-band frequencies, analyzing how LEO and GNSS measurements interact, and evaluating how the rapidly changing satellite geometry of a low-orbit constellation affects positioning performance. LEO satellites orbit at roughly 1,200 km compared to GNSS constellations at 20,000+ km, delivering significantly stronger signals and faster geometric diversity — properties that could dramatically accelerate convergence time for precise positioning and improve robustness in urban canyons and other challenging environments. u-blox’s X20 platform already supports all GNSS bands (L1, L2, L5, L6) with an integrated L-band receiver, making it a natural candidate for multi-layer signal fusion. ESA’s full Celeste demonstration constellation will ultimately comprise 11 satellites testing innovative signals, with the program endorsed as one of three pillars of Europe’s new Resilience from Space initiative.

u-blox Expands High-Precision GNSS with Galileo HAS and Infrastructure-Free PPP

u-blox announced the ZED-X20P-01B, a new variant of its all-band GNSS module that integrates Galileo High Accuracy Service (HAS) support — delivering globally accessible precise point positioning corrections broadcast directly from Galileo satellites without any dependency on local RTK base stations or internet connectivity. The module achieves reliable decimeter-level positioning anywhere on the planet using only sky-visible signals.

The ZED-X20P-01B also introduces moving-base functionality for precise relative positioning between autonomous platforms, enhanced jamming and spoofing detection verified at Jammertest 2025, and maintains the established ZED form factor for drop-in upgrades. Target applications span UAV mapping in connectivity-limited areas, marine dredging and seabed surveys, precision agriculture in remote regions, and robotic platforms requiring reliable relative positioning. Samples and evaluation kits will be available in June 2026, with live demonstrations scheduled at XPONENTIAL 2026.

Key Takeaways

AI compute spending is entering the $50 billion era: OpenAI’s disclosed annual budget dwarfs the GDP of many nations and signals that the infrastructure arms race is accelerating, not plateauing — with downstream pressure on chip supply, power grids, and data center capacity.
Supply chain diversification is now a boardroom priority: Apple’s quiet outreach to Intel and Samsung shows that even companies with the deepest supplier relationships are planning for a world where TSMC access cannot be guaranteed.
GNSS is going multi-layer: u-blox’s LEO-PNT integration work and infrastructure-free PPP via Galileo HAS signal a shift toward positioning systems that combine signals from multiple orbital altitudes and eliminate ground infrastructure dependencies.