Google's $811 billion in purchase commitments by Q2 2026 dwarfs Apple's flat $57 billion, per analyst Claus Aasholm. Combined hyperscaler obligations near $2 trillion mark a tectonic power shift from consumer electronics to AI infrastructure.
Key facts
- ~$2T: Combined hyperscaler commitments by Q2 2026
- $811B: Google's total commitments, up from ~$145B
- $678B: Microsoft's total obligations
- $57B: Apple's flat commitments, $56.2B due in 12 months
- $119B: Nvidia's commitments, exceeding Apple's
The AI infrastructure race has produced a new class of purchasing champions. Alphabet, Microsoft, Meta, and Amazon now hold roughly $2 trillion in total long-term purchase commitments by Q2 2026, according to analyst Claus Aasholm's estimates as reported by Tom's Hardware. The figure spans foundry capacity, 3D NAND, and DRAM, and should be taken with a grain of salt — but the direction is unmistakable.
Google is the most aggressive, jumping from roughly $140–150 billion in Q3 2025 to ~$811 billion by Q2 2026. Microsoft follows at ~$678 billion, Meta at ~$349.3 billion, and Amazon at ~$130 billion. These are total commitments, not memory-specific; the source does not break out the memory portion per company.
Apple, the largest memory consumer just a couple of years ago, sits at ~$57 billion — of which $56.2 billion is payable within 12 months. That trails Nvidia's $119 billion in commitments. The smartphone giant's flat curve against hyperscaler hockey sticks is the clearest signal that memory procurement power has migrated.
Key Takeaways
- Hyperscalers hold ~$2T in AI hardware commitments; Google leads at $811B while Apple trails at $57B.
- Memory becomes strategic asset.
Memory as a strategic weapon
The structural shift matters beyond the balance sheets. Memory has moved from commodity to strategic asset, and suppliers — Micron, Samsung, SK hynix — are gaining pricing power. Aasholm's estimates suggest these vendors will need major capacity expansion, though they've been disciplined so far. The commitments also feed directly into the AI supply chain: Google's ~$811 billion includes the TPU packaging deals with Intel previously reported, including the 3 million TPU packaging commitment by 2028.
The concentration is the story. A handful of CSPs now dictate terms that used to belong to Apple. If these commitments hold, memory pricing power shifts decisively to suppliers — and any hyperscaler pullback becomes a macro event, not a sector one.
What to watch
Watch the Q3 2026 earnings disclosures for actual memory purchase obligations versus total commitments, and whether Micron, Samsung, or SK hynix announce capacity expansions matching the ~$2T figure. A single hyperscaler guidance cut would test whether the pricing-power thesis holds.
Source: tomshardware.com
