Anthropic's IPO as a Turning Point
The second half of 2026 could beat the first half for AI investing. SpaceX went public a few months back, but Anthropic's IPO may end the year with a bigger bang. Anthropic is the gold standard for proprietary AI models, serving both individuals and enterprises.
The more important force is the arrival of open-source models. This echoes the internet era of the 1990s: first a wave of proprietary IPOs, then a much larger wave of open-source companies. Google, Salesforce, and Uber all came in that second wave. History will rhyme, if not repeat.
Cheaper, Safer AI Shifts Power to the Application Layer
Cheaper, more accessible, and safer AI unlocks venture capital and entrepreneurs. The market shifts from producers of AI to consumers of AI. Open-source models will spawn many types of companies - digital and physical AI applications, for both consumers and enterprises, across all large verticals.
The investment focus is the next layer, the application layer: how these technologies rebuild the largest industries.
Will the Investment Turn Into Real Profit?
Question: how will the application layer's profitability work, and can all this investment turn into sustainable profit?
Answer: LLMs laid the railroad tracks, and now the tracks can run, like earlier internet and technology cycles. AI is reimagining the largest workflows and customer problems, and that flows through to the bottom line. Investments span AI for commerce, insurance, cybersecurity, manufacturing, and space tech - all making the largest industries more efficient.
Risks to the AI Story
Question: as AI moves beyond hyperscalers into HR software, travel, commerce, and other sectors, is there any risk to that narrative?
Answer: there is much talk of AI doom, and any new technology takes time to figure out with problems to work through. Still, there is far more opportunity ahead, from company efficiency and on the labor side. AI will be a boon to labor. One portfolio company, Nextwork, upskills the next generation with AI and gives people jobs they never could have had - for example, a janitor who moved into a tech job after using the platform to upskill.
Cybersecurity: Good AI Against Bad AI
Cybersecurity is set to ramp as AI ramps. It is a volatile space within tech and an underappreciated part of the broader AI build - more technology needs more cybersecurity - yet it has not matched the boom seen in chipmakers.
There is bad AI and good AI. Every new technology draws bad actors, and you need good AI to fight bad AI. One portfolio company, Msubi, actively spots bad actors on large social networks - fake accounts, junk comments, and other garbage. Cybersecurity firms that clean up bad AI are clearly needed.
The Golden Age Is Still Ahead
The golden age of AI investing is still ahead. In the early 1990s, Sun Microsystems, Netscape, and Amazon came to market, and people thought the internet was spoken for. Open-source models and open weights sparked another wave of application-layer companies - Uber, Airbnb, Salesforce, and others. We are now in a golden age of entrepreneurship and investing, and a similar pattern will play out. Despite the common call that this is a massive bubble, it does not look that way in the current environment.
