← willsee.cc
AI Special Feature #1

The Global AI Compute Buildout

AI Special Feature #1 · July 27, 2026

Nine Deals, Four Continents

In the past 30 days, the world's largest technology companies committed over $1.15 trillion to AI infrastructure. Data centers. Chips. Power. Capital. Here are the nine deals — with their physical locations — that define the buildout.

1. Nvidia is in talks to backstop ~$250B for OpenAI's 10GW data center — the single largest compute financing ever. 📍 Abilene, Texas (Stargate campus) · US Southwest expansion for power
2. AMD signs a $5B compute-and-equity pact with Anthropic. Anthropic commits to 2GW of MI450 from H1 2027. First chipmaker equity check into a foundation-model lab. 📍 Anthropic primary clusters (AWS US-East/US-West) · AMD silicon from TSMC Phoenix, Arizona
3. TSMC raises 2026 capex to $60B–$64B (+15%). The world's largest contract chipmaker is betting AI demand isn't slowing down. 📍 Phoenix, Arizona · Kumamoto, Japan · Dresden, Germany (planned) · Tainan & Hsinchu, Taiwan
4. SK Group + Nvidia unveil a $500B+ AI initiative — HBM memory, next-gen chip packaging, and joint data centers exceeding 2 GW. 📍 South Korea (Icheon, Yongin cluster) · Nvidia $1B Naver AI center in Sejong, South Korea
5. Samsung lands a $200B+ contract to make chips for Broadcom through 2030 on 2nm and below. 📍 Pyeongtaek, South Korea · Taylor, Texas ($17B new fab)
6. Alphabet holds $94.1B in SpaceX equity. AI compute and space infrastructure — satellites, launch pads, Starlink ground stations — converging on the same balance sheet. 📍 Starbase, Texas · Cape Canaveral, Florida · Vandenberg, California · Starlink ground stations globally
7. South Korean FDI into the US surges 100%+ YoY to $10.2B in Q1 — driven almost entirely by AI infrastructure. 📍 Georgia (SK battery) · Kentucky · Tennessee · Indiana (SK Hynix HBM packaging)
8. SK Hynix completes a record $26.5B IPO — the largest in Korean history. AI memory demand is reshaping capital markets. 📍 Icheon & Cheongju, South Korea · West Lafayette, Indiana (new HBM facility under construction)
9. Microsoft, Alphabet, Meta, Amazon — Big Tech combined capex at all-time records. Data center construction consuming an unprecedented share of global industrial capacity. 📍 Microsoft: Phoenix AZ, Des Moines IA, Boydton VA, Gävle Sweden · Alphabet: The Dalles OR, Council Bluffs IA, Hamina Finland · Meta: Huntsville AL, Altoona IA, Odense Denmark · Amazon: Northern Virginia (world's largest cluster), Boardman OR, Columbus OH

12-Month Comparison: Then vs Now

July 2025
$0
No AMD-Anthropic compute pact
July 2026
$5B
First chipmaker × AI lab equity deal
July 2025
$28B
TSMC annual capex
July 2026
$64B
+129% — TSMC capex more than doubled
July 2025
~$50B
Total global AI data center market
July 2026
$1.15T
Single-deal commitments now exceed entire 2025 market
July 2025
No Nvidia-foundation-model financing
July 2026
$250B
Nvidia backstopping OpenAI — new category entirely
July 2025
$5.8B
SK Hynix market cap (pre-IPO estimates)
July 2026
$26.5B
Record Korean IPO — AI memory premium realized
July 2025
0
No 10GW single-site data centers planned
July 2026
1
OpenAI Stargate — first 10GW campus, Abilene TX

By the Numbers

Nvidia/OpenAI $250B
SK/Nvidia $500B+
Samsung/Broadcom $200B+
Alphabet (SpaceX) $94.1B
TSMC (capex) $64B
SK Hynix (IPO) $26.5B
SK FDI (US) $10.2B
AMD/Anthropic $5B

~$1.15 trillion committed in 30 days · 12 months ago, the entire global AI data center market was estimated at ~$50 billion.

One Observation

Twelve months ago, the AI compute buildout was measured in billions. Today it is measured in hundreds of billions — and individual data center campuses are now spec'd in gigawatts, not megawatts.

The numbers are so large ($250B here, $500B there) that they risk becoming abstract. But the physical facts are concrete: fabs are being dug in Arizona and Japan, steel is going up in Texas and Indiana, and the power purchase agreements are already signed.

This is not a valuation story. It is a construction story. The question is no longer whether the world will build AI infrastructure. The question is who pays for it, who powers it, and whether the companies that can't afford a seat at the table survive the next 24 months.

© 2026 WillSee · willsee.cc

Content is for informational and analytical purposes only — not investment, financial, or legal advice.