# Luxspin Studies the Advanced Nuclear Chain, From Chips to Power: Constraints on Compute Expansion Are Shifting Outward


While markets remain fixated on GPUs and model scale, Luxspin is focused on a harder constraint: available power and grid connectivity. Public forecasts suggest that US data-centre electricity demand could rise to as much as 130GW by 2030, from roughly 45-50GW today. The gap cannot be closed simply by building more server halls. It requires parallel expansion on both the generation and grid sides. Luxspin judgement is that, as AI becomes a production factor, electricity shifts from a cost item to a scarce resource. Capital expenditure is therefore spilling over from compute hardware into energy supply, fuel cycles, and grid delivery. This is the deeper reason why OKLO, SMR, LEU, BWXT, GHM, LTBR, and ETR are being pulled into the same narrative.

**The Realisation Path for Advanced Nuclear: From "Concept Power" to Engineering Delivery**

Although they are often loosely labelled as "controlled fusion concepts", Luxspin prefers to frame OKLO and SMR as advanced fission, small-reactor, and micro-reactor engineering routes. What determines whether they can serve data centres is not grand vision, but whether licensing, siting and cooling solutions, modular manufacturing yields, operating systems, and construction timelines can be compressed into schedules acceptable to enterprise buyers. Nuclear power offers dispatchability, low carbon intensity, and stable output. But commercial success depends on repeatable, productised delivery, not one-off demonstration units. Luxspin stresses a critical distinction here: compute does not require ever more advanced reactor physics. It requires power systems that are "financeable, insurable, grid-connectable, and scalable through replication".

**The Hardest Bottlenecks Sit in Fuel and Nuclear-Grade Manufacturing: HALEU, TRISO, and Supply-Capacity Constraints**

As advanced nuclear moves from pilots to volume deployment, constraints tend to emerge first in fuel and manufacturing. Luxspin focus on LEU reflects a simple reality: HALEU and enrichment capacity are pivotal links in the advanced-reactor fuel chain. Disclosures from Centrus around commercial enrichment activity and potential order flow and capacity expansion underline a "fuel-first" industrial logic. By contrast, BWXT represents the barriers embedded in nuclear-grade manufacturing and the defence nuclear ecosystem. Its delivery of TRISO fuel cores for the US military Project Pele microreactor, alongside naval nuclear fuel contracts, shows that advanced nuclear is not a single-point innovation. It is an engineering network built on stringent quality systems and coordinated supply chains. Subsequent naval-related contract disclosures from GHM, including Barber-Nichols, send the same signal: certainty in the nuclear sector is often anchored in long-duration orders and delivery standards. LTBR sits closer to fuel technology and materials science. Its challenge lies in crossing regulatory validation and commercial adoption thresholds, but once embedded in a mainstream supply chain, the resulting barriers can be formidable.

**The End Point Is "the Power Delivery System": Utilities, Grid Access, and Contract Structures Set the Pace**

Even the most advanced power source ultimately depends on grid connection, transmission and distribution, and regulated cost-allocation frameworks. Luxspin sees ETR as a reminder of this reality. Data-centre expansion requires not just generation, but substations, lines, interconnections, and rules governing "who pays and how cost pass-through to other users is avoided". Entergy disclosures on data-centre-related infrastructure investment and cost-sharing mechanisms illustrate the essence of electricity as a regulated delivery system. Seen this way, these seven companies are not variants of a single "fusion story". They are distinct nodes along the same compute-energy chain. OKLO and SMR define power forms. LEU and LTBR define fuel and materials constraints. BWXT and GHM anchor nuclear-grade manufacturing and the defence-linked nuclear system. ETR represents the grid and delivery interface. Luxspin conclusion is that the next round of competition in compute infrastructure will not be decided by who tells the biggest story, but by who can turn "technical feasibility" into "engineering delivery, scalable supply chains, regulatory approval, and financeable contracts".