Skip to content 
In today’s edition,
- Ads for scale
- AI timeline slips
- AI reshapes work
- Bitcoin ATMs rebound
Market Snapshot
Markets bounced back this week, lifted by easing U.S.-China trade tensions and softer inflation. The Nasdaq closed up 2.87% at 19,211.10, the S&P 500 gained 2.60% to 5,958.38, and the Dow added 1.80% to end at 42,654.74.
The big trigger was a 90-day suspension of tariffs between the U.S. and China, bringing duties on most goods down sharply, 145% to 30% on U.S. tariffs, and 125% to 10% on China’s. That helped set a positive tone early in the week, along with news that Saudi Arabia will now be allowed to buy advanced AI chips from U.S. firms.
Inflation data helped too. April’s CPI came in at 2.3% year-over-year, a bit below expectations, and wholesale prices actually fell, PPI dropped 0.5% from March. That gave markets more confidence the Fed will stay patient.
But not everything looked strong. Retail sales in April slowed to just 0.1%, and consumer sentiment dropped again, with more people worried about trade and inflation. Expectations for price increases jumped to 7.3% for the year ahead.
So while the week ended strong, the mood is still mixed with macro relief is here for now, but consumer caution is building.
Stock market closing data for the week of May 12th to May 16th, 2025
News Summaries
Data from the Ramp AI Index, which tracks billions in corporate spend across 30,000 U.S. companies, shows OpenAI leading AI adoption by a wide margin. But staying ahead in a market crowded with Google, Meta, xAI, and Anthropic requires more than paid subscriptions. To match rivals in compute, R&D, and user growth, OpenAI needs broader monetization. Ads offer that path, allowing the company to keep ChatGPT free for casual users (who are the majority) while generating the kind of high-margin revenue that historically fueled the rise of Google Search and Facebook. The recent hiring of Fidji Simo, who built ad platforms at Meta and Instacart, signals intent. If ads are kept separate from model outputs and transparently shown, they can fund growth without eroding trust. Even at just 1/10th of Meta’s ARPU, 1 billion users could mean $20 billion in annual revenue, most of which will likely be funneled back into compute. That’s a direct tailwind for Nvidia. More tokens, more usage, more GPUs. In the end, OpenAI isn’t just training models; it’s training habits.
Meta’s flagship AI model “Behemoth,” slated for an April release, has now been pushed to fall or later as engineers struggle to show clear gains over earlier Llama versions. Despite Meta’s public claims and a planned $72 billion in capex this year, much of it for AI, the model’s internal performance isn’t meeting expectations. Frustration is brewing at the top, and leadership changes in the AI group are being discussed. The delays come on the heels of Meta submitting a tweaked model to top a chatbot leaderboard, one that differed from the public release, raising concerns around transparency. With OpenAI and Anthropic also facing model setbacks, it’s becoming clear: pushing the frontier is getting slower, costlier, and less about hype, more about real, measurable breakthroughs.
Microsoft’s recent layoff of around 6,000 employees hit engineers hardest, with software developers making up over 40% of the 2,000 roles cut in Washington state alone, despite being the very people who build the company’s core products. As much as 30% of code on some projects is now written by AI, according to CEO Satya Nadella, reflecting how automation is starting to displace even high-skilled roles. Product managers and technical program managers made up another 30% of the cuts, while customer-facing roles like sales and marketing were largely unaffected. Though Microsoft framed the move as “delayering,” only 17% of those let go were managers, mirroring the broader company average, suggesting this was more about AI-driven reprioritization than org restructuring. As more companies like Salesforce and Workday also slow down engineering hiring while doubling down on AI, the message is becoming clearer: the AI shift isn’t just about new capabilities; it’s already reshaping who builds what, and how many are needed.
From the World of Crypto
Bitcoin Depot just posted a solid comeback.
After a $4.2 million loss last year, the company delivered a $12.2 million profit in Q1 2025. Revenue rose 19% year-over-year to $164.2 million, driven by more kiosk deployments and higher average transaction sizes. The market responded quickly as shares jumped 22% to $2.04. But that’s still far off from where they started; the stock is down 79% from its July 2023 Nasdaq debut.
The company, which runs over 8,400 Bitcoin ATMs across North America, is also leaning more into Bitcoin itself. It held $7.8 million worth of BTC in Q1, up from $600,000 the quarter before, after switching to fair value accounting.
CEO Brandon Mintz pointed to New York as the next big target. Bitcoin Depot doesn’t yet operate in New York but sees room for up to 3,000 kiosks. Talks with regulators are ongoing, and the firm hopes to enter the market in 2025.
That expansion comes with political heat. Regulators and lawmakers, especially Democrats, have flagged growing fraud concerns, particularly involving seniors. A bill now in motion would cap new-user transactions and mandate refunds for fraud reports. In 2024 alone, people over 60 reported $107 million in Bitcoin ATM-related losses. Bitcoin Depot says it’s responding with stronger compliance, dedicated monitoring, and 19 staff focused on fraud prevention.
Meanwhile, some lawmakers are leaning the other way. Rep. Lance Gooden recently suggested installing crypto ATMs in federal buildings to promote access and education. So the company sits at an interesting crossroads, growing again, showing profits, but navigating both regulatory pressure and public trust. A profitable quarter is a start, but the bigger challenge now is proving this isn’t just a bounce, but a business model that can last.
