Google to test custom AI chips in space on upcoming SpaceX mission

Investing.com -- Alphabet Inc's (NASDAQ: GOOGL) Google has officially launched Project Suncatcher, an ambitious orbital computing initiative aimed at deploying its proprietary artificial intelligence chips into space to bypass growing terrestrial energy and infrastructure constraints. The tech giant announced Thursday that it is preparing to launch a prototype satellite equipped with its Trillium Tensor Processing Units (TPUs) aboard SpaceX's (NASDAQ: SPCX) upcoming Transporter-18 rideshare mission, marking its first live operational test in orbit.

The initiative comes as hyperscalers face mounting land, permitting, and electrical grid bottlenecks on Earth, driving high-tech interest toward space-based cloud infrastructure. Industry momentum is already building around orbital data centers, underscored by SpaceX’s development of its Gigasat Factory in Bastrop, Texas, which is being designed to mass-produce AI-enabled satellites for deployment as early as late 2027.

To mitigate launch hazards and radiation risk, Google executed extensive ground testing, subjecting its Trillium hardware to high-g vibration simulation and proton beam bombardment at UC Davis’s Crocker Nuclear Laboratory. The tests indicated that the chips could survive a total ionizing radiation dose exceeding the requirements of a typical five-year orbital lifecycle, though engineers stress that live spatial deployment remains the ultimate proving ground.

Thermal management represents another primary technical hurdle, as the vacuum of space lacks airflow and forces hardware to rely entirely on heat pipes and radiative cooling systems to prevent TPU overheating. The upcoming mission will gather real-time telemetry on a new proprietary thermal deployment, providing crucial empirical data ahead of a secondary, multi-satellite interconnectivity test planned for 2027.

While commercial viability remains years away, Google’s measured, phase-gated approach highlights how tech conglomerates are rethinking long-term infrastructure scaling to sustain the massive compute demands of generative AI. Investors will likely monitor early mission telemetry and subsequent 2027 milestones to gauge whether orbital processing can evolve into a capital-efficient complement to traditional ground centers.

Original Article

You May Also Be Interested In

• BMO is bullish on this Magnificent 7 stock after enterprise AI agent launch
• UBS Picks Top Gold Mining Stocks as Sector Faces Cost Pressures in 2027
• S&P Global upgrades United Rentals to investment grade on leverage cushion