AI Code Flood, Both Sides
−7.0%
since Oct 8
Practiced to +2.6% over 4 days before going real.
Practice chapter: simulated results; no money was invested.
The idea
I believe AI coding assistants are now central to how software gets built, which means code ships faster but also messier. I want exposure to both sides: the companies providing the models and compute powering this shift, and the companies catching what AI writes through testing, observability, pipelines, and security review. The basket is concentrated in high-conviction names across both groups, no single stock dominates, and I'm holding some cash back to add on dips over the next year or two.
What it owns
- The companies selling the AI models and chips that coding assistants run on get paid every time an engineer leans on one. · NVIDIAIt manufactures the chips that power the leading AI coding models, and remains the standard choice as long as model training and inference demand keeps growing., MicrosoftIt owns GitHub and Copilot, sitting at the center where AI-written code gets made and shipped, and grows with paid AI coding seat adoption., AlphabetIt sells Gemini coding models and the cloud compute underneath, and remains relevant as long as Gemini is a serious choice for engineering teams., BroadcomIt builds custom AI chips for the big model companies, offering a second path to compute wins if hyperscalers keep ordering custom silicon to run coding models cheaper., Taiwan SemiconductorNearly every chip powering coding assistants gets manufactured by it, so it captures upside from AI compute demand growth regardless of which chip designer wins., Advanced Micro DevicesIt is the challenger bet on compute, and wins if running AI coding assistants at scale gets expensive enough that buyers want a second supplier with real inference chip orders.40%
- Sloppy AI-written code opens security holes, and someone has to scan and patch them. · Palo Alto NetworksIt scans code and cloud setups for the security holes sloppy AI code leaves behind, and wins as security review of AI-written code becomes standard practice., CrowdStrikeIt becomes the destination when sloppy code causes breaches, and benefits as AI-era security incidents push enterprises toward the breach response leader., SentinelOneIt uses AI to catch security problems at machine speed, matching the pace AI code ships at, and is the riskier pick on the security side that must keep winning share.17%
- Faster, sloppier code means more bugs reaching production, and these tools are how teams find them. · DatadogIt finds and alerts on failures when sloppy AI code breaks in production, and grows with observability spend tied to higher volumes of AI-written code., DynatraceIt watches production systems for the failures that happen when code ships faster and sloppier, and grows with monitoring budgets tied to AI-written code reaching enterprise stacks., ElasticIt is the log search layer teams dig through when AI-written code misbehaves, plus sells search tooling for AI apps themselves, and benefits as log volume grows.16%
- When AI writes more of the code, the platforms where code gets reviewed, merged, and shipped handle more volume. · JFrogIt tracks what is actually in the software being shipped, which becomes critical when AI writes code that pulls in packages nobody checked., GitLabIt is the pipeline the flood of AI code runs through for review, testing and deployment, and gains value as teams pay more to manage higher code volume., AtlassianIt is where engineering teams coordinate, and more code shipping faster means more tickets to track and coordinate across the team.14%
- Cash13%