Subsystem Roadmaps 2026 → 2030 (Consolidated)

The forward-looking sections scattered through Parts 4 through 9 each end with the same uneasy sentence: this subsystem is about to replatform. Power is moving off AC. Cooling is moving off air, and single-phase liquid is already being framed as the floor rather than the ceiling. Silicon is doubling memory and tripling fabric bandwidth on a yearly cadence. The network is fighting a three-way standards war while optics migrate onto the package. Storage is being pulled into the GPU's memory hierarchy. All five sit downstream of the same accelerator roadmap, which is why they move in sync — five faces of the power-bound density ramp. This chapter consolidates those forward pointers into a single 2026 → 2030 view and sorts the moves you can defer from the ones you must commit before steel is cut.

A roadmap is a register of option premiums. Betting on the wrong generation of GPU is reversible: you buy the next tray. The expensive mistake of the 2026 era is building a substrate that cannot absorb the generation after next, because the floor, the water, the risers, and the voltage class are the things you cannot retrofit cheaply once the hall is live. We walk the five subsystems in the order the cascade flows — power, cooling, compute/memory, network/optics, storage — and close on the rack and facility integration that ties them together, because by 2027 the unit of design stops being the 19-inch rack and becomes the ~600 kW power-and-cooling chassis.

The power chain is the subsystem under the most acute roadmap pressure, because it sits at the intersection of two exponentials: the per-rack power draw climbing from ~40 kW (H100, 2023) through ~120–142 kW (GB200/GB300, 2024–2025) to ~190–230 kW (Vera Rubin, 2026) and ~600 kW (Rubin Ultra Kyber, H2 2027), and the grid-side scarcity that made power the binding constraint in the first place (Chapter 16.1). At ~40 kW you can rectify AC inside the rack and not think about it. At ~600 kW you cannot: the copper cross-section, the rectifier count, and the I²R losses of pushing that power at 415/480 VAC to the rack become physically and economically untenable. The roadmap answer is a step-change in distribution voltage — and once the building commits to a voltage class, it stays committed.

The choice is between staying on the familiar 415/480 VAC chain with rack-level rectification — fine through roughly 150 kW — and taking the ±400/800 VDC disaggregated-sidecar path that NVIDIA, the OCP Mt Diablo project (Google/Meta/Microsoft), and the major electrical vendors (ABB, Eaton, Schneider, Vertiv) have converged on for the >150 kW rack. 800 VDC carries over 150% more power through the same copper than 415/480 VAC, and the end-to-end electrical chain efficiency rises from a representative ~82% AC path (the full AC span runs ~61–87.5%) to ~87% on an SST/800 VDC path — about a 5-point end-to-end gain that, at gigawatt scale, is tens of megawatts of recovered capacity. The catch is that the DC ecosystem is immature: the solid-state transformer (SST) that makes medium-voltage-to-800 VDC conversion efficient hit ~98% at 400 kW in an ETH Zurich benchmark and targets ~99%, but UL listing is not expected until roughly 2029. The decision you face in 2026 is therefore to design the building's voltage class and riser topology for the DC future while bridging with conventional gear until the SST and DC-breaker supply chain matures.

Underneath the distribution question sits the generation question. The interconnection wall has pushed behind-the-meter (BTM) on-site generation from a fringe tactic to a mainline strategy: ~90 GW of BTM gas has been announced cumulatively by mid-2026 (though only ~1 GW is under construction and ~2 GW online), because energizing megawatts on-site in 18–36 months beats waiting four-to-seven years in a utility large-load queue. So the power subsystem is no longer just a distribution-engineering problem inside the fence — it is a generation, fuel-supply, and grid-interactive problem, and the building must be designed to host the generation it may need to bridge to grid power. → DC architecture in Chapter 4.7; on-site generation in Chapter 4.8; the queue and speed-to-power in Chapter 3.2.

Cooling crossed its decisive fork earlier than the other subsystems, and the roadmap reflects a settled near-term and a contested far-term. The near-term is settled: single-phase direct-to-chip liquid cooling is the 2026 default, holding roughly 55% of the liquid-cooling market, because air saturates near 41 kW/rack and the GB200 NVL72 draws ~120–132 kW. Designing a new dense hall for anything but liquid in 2026 is a category error. The forward question is what comes after single-phase D2C when per-die thermal density keeps climbing toward and past 2 kW.

Here the roadmap forks into two candidate successors, shaped as much by chemistry and liability as by thermodynamics. Two-phase approaches (two-phase D2C and two-phase immersion) offer better heat transfer by exploiting latent heat of vaporization — but two-phase immersion stalled hard when the PFAS health-and-liability crisis drove 3M to exit the Novec fluorochemical business, leaving the supply chain and regulatory picture for the dielectric fluids deeply uncertain. Microfluidics — etching coolant channels directly into the silicon, as in Microsoft's research demonstrating up to ~3x cold-plate performance with AI-designed bio-inspired channels — is the more promising far-term path because it attacks the thermal resistance at its source (the die-to-coolant interface) rather than downstream of it. Neither is a 2026 production decision, so the conservative move is concrete: build for single-phase D2C now, but reserve the mechanical and facility-water headroom (CDU capacity, secondary-loop delta-T margin, manifold and quick-disconnect provisioning) so the hall can adopt the successor without re-plumbing. → DLC in Chapter 5.4; immersion and the PFAS problem in Chapter 5.5; facility water loops in Chapter 5.7.

The accelerator roadmap sets the clock for every other subsystem in the building. The publicly-stated cadence is yearly: Vera Rubin (NVL72, ~190–230 kW rack) in 2026, Rubin Ultra in H2 2027 — deployed either as the NVL576 system (eight MGX NVL72-class racks in one 576-GPU optical NVLink domain) or in the ~600 kW Kyber rack (NVL144: 144 quad-die packages = 576 GPU dies; eight Kyber racks form NVL1152) (as shown at GTC 2026; supply-chain reports from Apr–Jun 2026 say Rubin Ultra was scaled back to a dual-die 2+2 Kyber layout over CoWoS-L warpage, unconfirmed by NVIDIA, so per-package specs are in flux), ~15 ExaFLOPS FP4 inference / 5 ExaFLOPS FP8 training per rack — and Feynman in 2028, widely expected to push toward the 1 MW rack. Memory climbs in lockstep: per-GPU HBM goes H100 80GB → B200 192GB → B300 288GB → Rubin 288GB (HBM4) → Rubin Ultra 1TB (HBM4e). For a strategist the takeaway is the cadence, not the spec sheet. A yearly generation step with a roughly 2–3 year frontier-economic life (the contested depreciation figure from Chapter 1.8) means the substrate you pour in 2026 must absorb at least two generations of density growth within its first refresh window.

On first live silicon, CoreWeave measured a Vera Rubin NVL72 delivering up to ~10× the tokens per megawatt of a GB200 NVL72 (DeepSeek-R1 at a matched interactivity target), which NVIDIA pairs with one-tenth the inference cost per million tokens, up to 10× the inference per watt, and one-fourth the number of GPUs (~75% fewer) to train a given MoE model — a training metric. These are vendor ratios versus Blackwell, not absolute $/token: no absolute cost-per-token or tokens-per-dollar has been published.

The roadmap risk on this axis is the advanced-packaging gate, not the GPU. The binding constraint above wafer fabrication is TSMC CoWoS capacity and HBM stacking, not transistors. 2026 HBM3E is effectively sold out with a ~30% supply gap and ~15–20% per-quarter price rises, and HBM4 (qualified with Samsung and SK hynix for Rubin) inherits the same hybrid-bonding and CoWoS-wafer bottleneck. The consequence for a data-center program is counterintuitive: your delivery schedule is gated less by your capex and more by an upstream packaging line you do not control. A scope that assumes GPU availability on the vendor's stated cadence, without an allocation agreement, is underwriting a schedule it cannot guarantee. → NVIDIA roadmap in Chapter 7.2; HBM in Chapter 7.6; advanced packaging in Chapter 7.7.

The network roadmap is where the most consequential standards bet lives, because unlike power or cooling — where physics largely dictates the answer — the fabric question is partly a political one about which ecosystem you tie your accelerator optionality to. Two distinct fabrics are evolving in parallel. Scale-up (intra-rack, the memory-semantic domain that binds a tray of GPUs into one logical accelerator) is the contested ground: NVIDIA's NVLink (1.8 TB/s/GPU on Gen5, 3.6 TB/s on Gen6) with NVLink Fusion opening the IP to third-party CPUs/XPUs; the open UALink / UALoE consortium path (AMD-led, up to 1,024 accelerators in the 1.0 spec); and Broadcom's SUE/ESUN (Scale-Up Ethernet). Scale-out (inter-rack, the all-reduce fabric) is consolidating faster around Ultra Ethernet (UEC 1.0: packet spray with reorder, UCCM congestion control, native RDMA) as the open answer to InfiniBand, alongside NVIDIA's Spectrum-X.

The scale-up standard you underwrite is a multi-year lock-in decision: it sets which accelerators you can rack, how large a scale-up domain you can build (and therefore your tensor-/expert-parallel ceiling), and your copper-vs-optics reach budget. Choose NVLink and you get the largest, most mature scale-up domains and the deepest software stack — at the price of single-vendor accelerator dependence. Choose UALink or SUE and you preserve multi-vendor optionality at the price of a younger ecosystem and smaller proven domains. Underneath both sits the optics transition: as lane rates climb to 200G and 448G and reach budgets shrink (passive copper ~1–2 m at 800G/1.6T), co-packaged optics moves the optical engine onto the switch package — cutting per-1.6T-link power from ~30W (pluggable DSP) to ~9W and delivering ~3.5x energy efficiency with 10x resilience. NVIDIA's Quantum-X (InfiniBand) reached availability in early 2026 and Spectrum-X Photonics (Ethernet) follows in H2 2026. If your 2027 build uses a ~600 kW Kyber rack, rack-to-rack optics and CPO are very likely on your critical path, not an option. → scale-out standards in Chapter 8.4; topology and oversubscription in Chapter 8.5; CPO and fiber plant in Chapter 8.10.

Storage is the subsystem most often left off a roadmap, and the most quietly transformed by the density ramp. The forward direction has four threads. First, the data path is moving off the CPU: GPUDirect Storage and GPU/DPU-initiated I/O let the accelerator pull data over NVMe-oF without a host-CPU bounce, and the new generation of DPUs (NVIDIA BlueField-4 at 800 Gb/s) is built to own this path. Second, the spinning disk is being designed out of the hot path entirely — all-flash is becoming the default for both the training scratch tier and the checkpoint tier, with PCIe 6.0 dense-flash servers (e.g. 96 E3.S SSDs, ~2.9 PB) emerging to feed it. Third, the historical split between parallel file systems and object stores is converging, as the same platforms (WEKA, VAST, DDN) serve file and object semantics over one flash substrate. Fourth, and most distinctively for inference, CXL and Ethernet-attached flash are becoming a KV-cache tier — a new layer of the memory hierarchy that sits between HBM and bulk storage.

That last thread is the one with the sharpest 2026 roadmap signal. As reasoning models emit long decode sequences, the KV cache balloons, and the economics of holding it in HBM collapse. The answer is a three-tier KV hierarchy — HBM, then host/CXL memory, then NVMe and Ethernet-attached flash — that NVIDIA is standardizing via BlueField-4's context-memory platform (CMX) and the NIXL transfer library, with vendors reporting the ability to serve roughly 10x more users by offloading prefix caches to flash. For the facility, storage is no longer a back-of-house capacity question; it is part of the inference memory hierarchy and sits on the latency-critical path. A 2026 design that treats storage as bulk capacity, decoupled from the fabric and the accelerator, is designing for the training workload of 2023, not the inference workload of 2026. → the CPU-bypass data path in Chapter 9.3; inference and KV-cache storage in Chapter 9.7; object/capacity tier in Chapter 9.6.

The five subsystem roadmaps do not converge on a server — they converge on a chassis. The defining structural shift of 2026 → 2030 is that the unit of design stops being the 19-inch rack populated with independent servers and becomes an integrated power-and-cooling enclosure where compute, NVLink switching, liquid distribution, and power delivery are co-engineered as one mechanical object. The GB200 NVL72 (~120–132 kW, ~1.36 t / 3,000 lb, 5,184 in-rack copper NVLink cables, blind-mate liquid manifolds, paired 1,400 A busbars carrying ~2,800 A aggregate at 48 V) is the first generation that can only be understood as a single integrated unit. The ~600 kW Kyber rack (NVL144) is the next: at that power the disaggregated sidecar (Mt Diablo / Diablo 400) moves power conversion out of the compute rack into an adjacent power chassis, the busbar becomes liquid-cooled, and scale-up optics move onto the package.

For facility design this collapses several previously-independent decisions into one substrate commitment. A ~600 kW chassis — no vendor mass figure is published, but the ~132 kW NVL72 already weighs ~3,000–3,300 lb wet, so Kyber-generation racks will land substantially heavier — implies a structural slab and seismic-anchoring basis you cannot retrofit (Chapter 6.2, Chapter 6.7); an 800 VDC riser and busway you cannot rip out without a hall outage; a facility-water capacity and pipe-rack geometry sized for liquid heat rejection at that density; and a row-and-aisle pitch set by the chassis footprint plus its sidecar. The same rule applies here as across the whole roadmap: reserve the irreversible substrate for the chassis you will house in 2027–2028, and keep the reversible fit-out — the actual trays, transceivers, CDUs, and storage shelves — matched to the generation you are buying this year. → rack as integration unit in Chapter 7.13; modular/prefab construction in Chapter 6.4.