12 predictions for 2026

Maybe it’s a little early for 2026 predictions and maybe these scream ‘perma-bull’ but I wanted to have some fun with tech and macro predictions for the coming year. These cover themes I’m particularly interested in, and the practice of making (and trying to quantify) somewhat loose predictions sharpens focus. There are plenty of bad tail risk things that could happen - Taiwan escalation, quantum cryptographic heist, bioterrorism - but I’ll shy away from that here :)

1/ The LLM ads business turns on in a meaningful way. OpenAI surges past $1B in ads revenue within 6 months. The total market could be over $5B in year one.

2/ METR task length eval trendline holds. Frontier models reliably complete a workday (~8 hours) of coding tasks 50% of the time.

3/ AI coding market continues to explode. Cursor hits $5B revenue run rate, Claude Code hits $15B revenue run rate.

4/ Perp volumes cross $10T one month. Drivers include more liquidity, RWA, clients / front doors, and potentially even new exchanges / trading platforms. Prediction market volumes (primarily Kalshi and Polymarket) exceed $30B in one month with more institutional flows and enterprise participation.

5/ Waymo crosses 1B rider-only miles. They continue their steady rollout across the United States and likely hit every “NFL city” by the end of 2027.

6/ OpenAI and Anduril go public. Both have to rationalize their expenses to some public shareholders, but I have little doubt there will be plenty of demand for the shares. I don’t think Stripe will go public.

7/ Microsoft buys Glean and Cognition. Satya is hungry for top flight product and engineering talent. Microsoft has existing product surface here, but could use upgrades.

8/ 5K home robots (with a predominantly U.S. supply chain) ship to consumers in H2. The battle to become “Apple of humanoids” kicks into full gear. The robotics supply chain (motors, actuators, sensors) becomes the equivalent of the AI capex trade (power, memory, cooling, optics) in the public markets.

9/ Zero shot antibody design means binder cost asymptotes to ~0. The universe of potential target proteins radically increases and more models shift to incorporate developability attributes.

10/ GLP-1 revenues cross $75B by end of year, showing plateauing growth and pricing pressure. Delivery modalities diffuse and insurers continue to process how to cover them. They still mark one of the biggest drivers of change to American consumer spending habits since the advent of the automobile, consumer packaged goods, and big box retail chains.

11/ Unemployment rate goes over 5%. AI effects start to kick in.

12/ Trump unveils an “Operation Warp Speed” effort (federal subsidies, deregulation, public-private partnership) for housing ahead of midterms. His actions will be seen as extreme and overreach, but the crux of the issue is bipartisan in nature.