AI Clears the Tech Debt
Royalbeck.design@royalbeckdesign
−2.2%
since Oct 16
The idea
I believe AI coding agents are about to clear out most of the tech debt that's been slowing software down, and this basket is built to ride that over years, not as a quick trade. It owns the companies building the cleanup agents, the developer platforms where that work actually runs, the chips powering it all, and a contrarian slice of legacy-heavy banks and insurers that stand to gain most as their old code gets cheaper to maintain. Positions are steadier in the big incumbents, smaller in faster-moving developer tools, with 10% held in cash to add if prices drop.
What it owns
- These companies sell the AI coding agents and modernization services that do the actual tech debt cleanup. · MicrosoftIt owns GitHub Copilot, one of the most widely used AI coding agents, so if agents chew through old code fast, much of that work runs on its tools and Azure., AmazonAWS sells agents built to migrate and rewrite legacy apps, and every rewritten system tends to land on its cloud., IBMIt is most tied to mainframes and COBOL, the oldest code banks and insurers still run, and it sells AI tools to modernize exactly that., AlphabetGemini coding agents and Google Cloud compete directly for the job of rewriting old code and landing migrated workloads on its infrastructure., ServiceNowIt sits on top of old systems and automates the workflows around them, so companies cleaning up legacy code tend to route that work through it., Palantir TechnologiesIt pulls data out of messy old systems and puts AI on top, which is a shortcut around tech debt for big organizations including banks.36%
- Code gets cleaned up, tested, shipped and moved onto modern systems through these tools, so more cleanup means more use. · MongoDBWhen old apps get rebuilt, many move off decades-old databases onto modern ones like it, picking up new workloads from modernization projects., SnowflakeIt is a common landing spot when companies pull data out of legacy warehouses and systems during modernization., GitLabIt runs the pipeline where code gets reviewed, tested and shipped, so a flood of AI-rewritten code has to pass through platforms like it., DatadogIt watches running software, and rewritten or migrated systems need to be monitored to prove they work., AtlassianJira and related tools are where engineering teams plan and track cleanup work, so more modernization projects means more use of its products., JFrogIt manages the packages and releases that every software change goes through, so faster code rewriting means more volume for its plumbing.24%
- Every agent rewriting old code runs on chips, so heavy cleanup work turns into chip demand. · NVIDIACoding agents run on a lot of computing power, and its chips power most of the data center compute that agents use., Taiwan SemiconductorIt manufactures most of the advanced chips that agents run on, so AI chip orders keep filling its factories., BroadcomIt builds custom AI chips for the big clouds, the same companies selling the cleanup agents and running the modernization work.15%
- Banks and insurers carry decades of old code, so they save the most money if AI really clears it. · CitigroupIt carries some of the heaviest legacy technology in banking, so it has the most to gain if AI makes that cleanup cheap and fast., JPMorgan ChaseIt spends more on technology than almost any bank, much of it keeping old systems alive, so AI that clears that debt would lower its costs., TravelersIt operates on legacy systems where AI-driven modernization can reduce the expense ratio, the share of premiums consumed by operating costs.15%
- Cash10%