Karsten Wenzlaff, Advisor
August 26th, 2025
Karsten Wenzlaff is a Crowdfunding and Fintech Expert from Germany. He is Secretary-General of Digit[...]
Scotiabank Joins US$22B AI Chip Financing
September 18, 2026 | NCFA Market Activity | Artificial Intelligence And Data, Capital Markets Infrastructure And Funding, Banking And Financial Institutions
Ten Banks Finance Google TPUs and Customer Contracts
On September 16, 2026, a group of ten banks including Scotiabank agreed to provide a US$22 billion AI infrastructure loan to Crux AI, a new U.S. based cloud venture backed by Blackstone and Alphabet. Goldman Sachs, Sumitomo Mitsui Banking Corp., Barclays and BNP Paribas are also among the lenders, according to Reuters. The US$22 billion applies to the full lending group rather than Scotiabank individually. Scotiabank's own commitment and economics haven't been disclosed, and the bank didn't comment publicly on the transaction.
The financing will fund purchases of Google's Tensor Processing Units (TPUs) specialized processors used to train and run AI models. The debt is backed by the processors and Crux AI customer contracts, giving lenders support from both the hardware and revenue tied to demand for its computing capacity. Both can change quickly as newer processors enter the market and customers reassess how much compute they need.
See: BGC Launches Compute Infrastructure Markets
The lenders are also working to bring additional banks into the loan through syndication, which would spread the exposure across a larger group. Several banks are reportedly providing a separate US$1 billion revolving credit facility, while the main debt could later be refinanced through the investment grade bond market. That would allow more of the financing to move from bank balance sheets to institutional investors.
AI Chips Are Becoming Financeable Infrastructure
Crux AI combines large amounts of equity, debt and specialized hardware in one infrastructure business. When Blackstone and Google announced the venture in May, Blackstone committed an initial US$5 billion in equity. Google is supplying TPUs, software and services, while Crux expects to bring its first 500 MW of capacity online in 2027. Customers will be able to buy data centre capacity, networking, operations and access to Google TPUs without building the facilities or buying large quantities of processors themselves.
The US$22 billion loan shows that banks are willing to finance AI processors against both the equipment and contracted demand for the computing capacity they provide. Expensive chips can therefore support large infrastructure loans when lenders are comfortable with the hardware, customer contracts and expected demand. In June, BGC launched compute infrastructure markets for trading, price discovery and risk management around compute and memory capacity. The Crux deal adds bank credit to a market that is already developing ways to trade and price computing capacity.
The Collateral Comes With Technology Risk
Traditional infrastructure lenders can estimate the useful life and resale value of assets such as buildings, power equipment or transportation fleets. AI processors are harder to value over long periods because a chip can remain technically useful while losing economic value quickly when a newer generation delivers more computing power at a lower operating cost. Customer contracts can reduce that risk because lenders aren't relying on hardware value alone, but those contracts still depend on customer quality, pricing, duration and continued demand for the underlying compute.
See: Goldman Sachs Buys Québec AI Compute Platform QScale
Bringing more banks into the facility spreads the exposure across a larger lending group. If some of the debt is later refinanced in the investment grade bond market, pension funds, insurers and other institutional investors could also gain exposure to AI infrastructure. That would widen the pool of capital available for new data centres and processors while spreading technology and credit risk across more investors.
Scotiabank's participation puts a Canadian bank inside one of the largest reported financings built around AI compute. It does show that global banks are prepared to finance AI hardware and contracted compute demand at a scale once associated with traditional infrastructure projects.
Talking Point
If AI processors and compute contracts can support tens of billions of dollars in bank debt, what happens to that financing model when a new generation of chips changes the value and economics of the collateral before the loans mature?
The National Crowdfunding & Fintech Association (NCFA Canada) is a financial innovation ecosystem that provides education, market intelligence, industry stewardship, networking and funding opportunities and services to thousands of community members and works closely with industry, government, partners and affiliates to create a vibrant and innovative fintech and funding industry in Canada. Decentralized and distributed, NCFA is engaged with global stakeholders and helps incubate projects and investment in fintech, alternative finance, crowdfunding, peer-to-peer finance, payments, digital assets and tokens, artificial intelligence, blockchain, cryptocurrency, regtech, and insurtech sectors. Join Canada's Fintech & Funding Community today FREE! Or become a contributing member and get perks. For more information, please visit: www.ncfacanada.org
 |  |  |
|---|---|---|
 |  |  |
Leave a Reply