Goldman Sachs revises its U.S. data center construction forecast upward amid headwinds, projecting that related spending will reach $125 billion by 2028.

Jefferies strategist Christopher Wood’s latest report states that over the four quarters ending in Q3 2026, U.S. AI-related real capital expenditure contributed 1.39 percentage points to year-over-year GDP growth, accounting for 67% of the country’s total 2.1% economic expansion. Meanwhile, investment in information processing equipment and software rose to 5.35% of nominal GDP, surpassing the 4.46% recorded during the 2000 dot-com bubble to hit an all-time high. Jefferies notes that widening credit spreads, ongoing declines in semiconductor stocks, and deteriorating free cash flow among hyperscale cloud service providers all signal the AI capital expenditure cycle may be nearing a turning point. Microsoft, Amazon, Alphabet, and Meta are projected to combine for roughly $1.04 trillion in capital expenditure this year, but Amazon and Alphabet’s free cash flow have fallen to -$35 billion and -$90 billion respectively. The report also warns that the AI ecosystem has built a $2.5 trillion backlog of contracts and a revolving financing system, with data center financing increasingly reliant on private credit and off-balance-sheet structures. A downward revision to AI investment expectations could trigger a chain reaction of tighter financing and shrinking capital expenditure. Jefferies argues that a sharp contraction in AI investment would deliver a significant shock to U.S. economic growth and may force the Federal Reserve to pivot to loose monetary policy. However, with the U.S. fiscal deficit already at 6.2% of GDP, U.S. Treasuries may not function as traditional safe-haven assets in that scenario. The report advises investors to properly diversify their asset allocations beyond AI, focusing on sectors such as gold and biotech.

AI data center developer Crusoe confirmed yesterday it has sold an unfinished data center project in southeastern Wyoming, U.S. to Google. Google plans to invest $17 billion to take over and continue development, though neither party disclosed the specific acquisition price. The project was first announced in 2025 by Crusoe and Tallgrass, an infrastructure firm backed by Blackstone, with a planned final expansion to 10 gigawatts of power capacity. Previously, Google had considered becoming a customer of the project, but ultimately opted to take over development due to concerns over construction costs and timelines under Crusoe’s management. The deal underscores the challenges independent data center developers face in large-scale AI infrastructure construction, including equipment procurement, power supply, cost control, and community opposition, among other issues. Crusoe has not ended its partnership with Google as a result; it is currently developing another data center near Amarillo, Texas for the tech giant. The company has previously built data centers for Oracle, OpenAI, Microsoft, and Meta, among others, and closed a $3.9 billion funding round in September this year, at a valuation of roughly $30.9 billion.

According to monitoring by Ai Yi, the address 0xBA2…aB9ff has accumulated purchases of 740,000 LINK tokens (valued at around $8.23 million) and 3,046.68 ETH tokens (valued at approximately $6.92 million) over the past five days, with a total investment of roughly $15.15 million. Currently, the address holds 787,000 LINK tokens. After recent position additions, its average cost basis for LINK has dropped to around $11.24, though it still faces an unrealized loss of approximately $366,000.

According to monitoring by on-chain analytics firm Lookonchain, whale address 0xC374 purchased an additional 308,517 UNI tokens approximately 10 hours ago, valued at around $2.13 million. Over the past two months, the address has accumulated a total of 1.32 million UNI tokens, with a total value of roughly $11.05 million.

Whale 0xC374 bought another 308,517 $UNI ($2.13M) 10 hours ago. Over the past 45 days, the whale has bought a total of 1.32M $UNI ($11.05M).

According to TradingBeats data, since January 1, 2026, Russian traffic has accounted for 10.55% of MEXC’s total platform traffic, making it the platform’s largest traffic source region; Iran accounts for 4.86%, ranking third. This data provides context for recent actions taken by the U.S. Treasury Department against MEXC. The Treasury’s Financial Crimes Enforcement Network (FinCEN) issued an order yesterday prohibiting U.S. financial institutions from processing fund transfers involving MEXC. FinCEN alleges that between May 2021 and May 2026, the platform processed at least $160 billion in funds linked to illegal activities including sanctions evasion and cybercrime, including nearly $1.5 billion in transactions related to Russian sanctions evasion and approximately $10 billion in direct transactions involving Iran. The U.S. action is based on Section 9714(a) of the Countering Russian Money Laundering Act, not traditional OFAC asset freeze sanctions. The relevant ban will take effect 30 days after the order is published in the Federal Register.