84% Of Small Businesses Are Now Using AI This Quarter. Here's Why That Locks In The Hyperscaler Trade.
The Journal's Tuesday A1 feature on Helene Godin's By the Way Bakery — a 14-person gluten-and-dairy-free shop in Pleasantville, NY — quietly contained the most important AI statistic of the week.
According to the survey cited in the piece, 84% of businesses with 20 to 99 employees and 91% of mid-sized firms plan to use AI this quarter. Among the smallest businesses (under 20 employees), the same survey found majority adoption is now the rule, not the exception.
The By The Way Case Study
By the Way previously tried Enterprise Resource Planning software — the SAP / Oracle / NetSuite class of system that runs Fortune 500 supply chains. "It was too hard to follow, and it wasn't reliable," said Godin. The cost was prohibitive and the maintenance overhead killed the deal.
What's working now: an AI agent built by a consulting firm called Streamliners. The agent scrapes daily order data, forecasts supply, syncs to the bakery's point-of-sale, and reorders ingredients automatically. Cost: a few hundred dollars per month, plus a few dollars in AI credits.
The CTO at Streamliners, Tanner De Jonge, framed it cleanly: "The bigger they get, the more important the data is." AI made data-driven operations economically viable for a business that previously could not afford it.
Why This Matters For The Hyperscaler Trade
The "AI bust" thesis has always argued that hyperscaler capex (Microsoft, Alphabet, Amazon, Meta) is running far ahead of actual demand. The Pope's encyclical, the chatbot-psychosis stories, and the data-center protests give that thesis some narrative momentum.
The 84% small-business adoption number breaks the bust thesis at the foundation. Here's why:
- The U.S. has roughly 33 million small businesses. If even 50% are running AI on hyperscaler cloud infrastructure within 24 months, that's 16+ million new compute customers.
- SMB compute is "sticky" in a way enterprise compute isn't. Small businesses don't have IT departments. Once their inventory or accounts-receivable system is on an AI agent, switching costs are enormous.
- The revenue per small business is small but the count is huge. Microsoft (MSFT), Alphabet (GOOGL), and Amazon (AMZN) don't talk much about SMB adoption on earnings calls because each customer is too small to single out — but aggregated, it pulls cloud consumption growth forward by years.
The Honest Counterweight
De Jonge's own quote in the article: "AI still has hallucinations. You have to build guardrails in so that it's not making mistakes." The technology is good enough to deploy in production but not good enough to deploy unmonitored. That's why a consulting layer (Streamliners and dozens like it) is going to capture a disproportionate slice of the SMB-AI value chain.
The risk: a single large hallucination at a major customer triggers a confidence reset. We saw a small version of this in late 2024 after the first wave of large-language-model embarrassments. The market shrugged it off. The next reset will likely also be shrugged off — but it's a real path to a 10-15% drawdown in the AI-cloud names without a change in long-term thesis.
For Retirees Reading This
If you have a target-date 2030 or 2035 fund in your 403(b) / 401(k), check the top-10 holdings. Most are 7-10% Microsoft, Alphabet, Nvidia, Amazon. That allocation is the right shape for the SMB-AI adoption cycle. The risk is concentration — not direction.
What we recommend for clients in CalSTRS / CalPERS supplemental accounts: hold the broad-market index for the AI exposure, and complement it with a defense / gold / dividend sleeve that doesn't move with hyperscaler regulatory headlines.