Atreides Management Leans Harder Into AI Plumbing And Profitable Software

The biggest buys are tightly clustered around one idea: own the bottlenecks and workflows created by AI, not just the compute itself.

On the hardware side, Atreides is clearly building an ecosystem around data movement. Astera Labs and new position Credo are both high‑conviction bets that the explosion of GPU deployments will stress interconnect and SerDes more than CPUs. A smaller but notable add to Micron reinforces the view that memory bandwidth and HBM supply will accrue outsized economics as AI training normalizes.

On the software side, the fund is willing to embrace controversy where it sees a path to durable cash flows. Unity’s near‑doubling, a 64.3% increase in Wayfair, and a 255.2% ramp in Roblox are all examples of leaning into battered platforms with large user bases and clear operating‑leverage upside if management execution improves.

The new positions in Zoom, Palo Alto Networks, HubSpot, and Synopsys round out the theme. Zoom is being re‑underwritten as a productivity and communications layer with AI monetization potential, Palo Alto as the security toll‑booth for distributed workloads, HubSpot as a high‑quality SMB software compounder bought after a drawdown, and Synopsys as a picks‑and‑shovels play on chip design and verification.

Even outside pure tech, the new stake in Vistra slots neatly into the thesis. More datacenters and AI training mean more power; Atreides is willing to own a utility with that structural tailwind despite it being temporarily below their average cost.

The sales this quarter look more like profit‑taking and quality rotation than a change of heart on AI or growth. The manager is freeing capital from big historical winners and marginal stories to fund newer, more asymmetric setups.

The most obvious funding sources are in optical and networking. Ciena was cut by 52.6% and Lumentum by 30.2%, both after enormous gains — Ciena is sitting over 625.3% above average cost and Lumentum an eye‑popping 1710.9%. Coherent was slashed by 48.9% with a 244.9% gain versus cost. These are classic “sell your old heroes to buy the new ones” trades.

Nvidia — once the canonical AI trade — was trimmed by 35.1%. Atreides is not abandoning it (it remains a 5.31% position) but is clearly reallocating away from a crowded winner toward less consensus infrastructure names like Astera Labs, Credo, and GlobalFoundries.

In software, the pattern is sharper. Snowflake was cut by 69.8%, with WIX and Compass both nearly halved, after disappointing return profiles relative to other opportunities. These are not panic exits — some are still above cost — but they signal cooling conviction in high‑multiple growth where unit economics or competitive position are less clear.

Consumer and services names like Wingstop and Dick’s saw trims too, which look like sources of liquidity rather than a macro call. By contrast, Atreides tolerated drawdowns in positions it likes (e.g., HubSpot, Affirm, ACV Auctions) only when that capital could be recycled into higher‑conviction infrastructure or platform software.

The sector bars would suggest a simple headline — Technology up, Utilities and Telecom down — but the underlying move is more nuanced: Atreides is upgrading its tech mix toward mission‑critical infrastructure and durable platforms.

Tech exposure jumped from an estimated 52.4% to 58.9% of the book. However, within that, capital is migrating from legacy optical and generic cloud data platforms into semiconductors, design tools, and software tied directly to AI workloads and network effects. The new or enlarged stakes in Astera Labs, Credo, Micron, Synopsys, and GlobalFoundries, plus incremental Intel, all push the book toward the physical constraints of AI: bandwidth, memory, and manufacturing.

Telecom and utilities on the surface look reduced, but that is mostly Ciena and Lumentum shrinkage. The new Vistra position means the fund is swapping stock‑specific optical risk for a more macro power‑demand bet, consistent with rising datacenter energy needs.

Real “Financials” exposure is modest but edging higher via Rocket, Chime, and a trimmed Affirm, all consumer‑credit or fintech names that can benefit from normalization and operating leverage. The sector label noise in the data (e.g., Mastercard and Visa tagged as Real Estate) obscures that Atreides is actually creeping up in payments and transaction rails, which rhyme with the broader theme of owning toll‑booths on digital activity.

Consumer Discretionary has nudged up, but the composition is skewed to idiosyncratic growth (Amazon, Wayfair, EchoStar, Wingstop) rather than broad retail beta — consistent with a manager still focused on stock‑picking over macro calls.

Taken together, the 13F reads like a clear view of where Atreides thinks the puck is going in AI: away from obvious GPU scarcity and into the messy middle — bandwidth, workflows, and security. The fund is willing to pay for quality (HubSpot, Synopsys, Palo Alto) while simultaneously scooping up bruised platforms (Unity, Zoom, Roblox, Wayfair, Rocket) where product‑market fit is proven but sentiment is not.

The book is more concentrated than a typical long‑only portfolio, with 53.4% in the top 10, but it is diversified across layers of the same secular themes. If the AI build‑out continues, Atreides owns the connectivity, memory, design tools, security, and power that must scale with it. If adoption slows, many of these software and payments names can still compound on more traditional drivers.

The main risk in the positioning is obvious: heavy tech and consumer‑growth exposure into a choppy tape. A -3.35% quarter did not prompt a risk‑off pivot, so drawdown tolerance is high. That said, the manager has shown a consistent willingness to harvest big winners and cut weaker stories to fund new ideas, which should give them dry powder if volatility spikes.

For outside observers, the signal is that Atreides believes the easy part of the AI trade is over. The next phase is about picking the overlooked plumbing and platforms that turn compute into cash flows — and this portfolio is already positioned as if that thesis is right.