Servers Direct (build-to-order AI infrastructure reseller)

NEGATIVE, DELIBERATELY NOT STRUCTURAL: no lead time in weeks is printed for these SKUs. It is not that no vendor in the chain publishes one - CDW does, and it is recorded on cdw-enterprise as 'published' - but this channel routes to 'request a quote' or shows a stock count instead, so the number exists elsewhere in the market and is not surfaced here. Checked 2026-10-04 per source spl-negative-no-oem-publishes-lead-time.

EAR 3A090.a / 3A090.b (advanced-computing ICs) + EAR 742.6(a)(6)(iii)(A) (D:5/Macau end-user and ultimate-parent) + EAR 744.23 (Entity List screening) + EAR 744.6 (Unverified/MEU lists) (or EAR 3A001 or EAR99 (below the 3A090.a TPP / performance-density thresholds) + EAR 744.23 (Entity List) + EAR 744.6 (Unverified / MEU lists), depending on the configured GPU)

MECHANISM: build-to-order with no named accelerator, so control status is a FUNCTION OF THE CONFIGURED PART, not of the channel. A box carrying an H100/H200/B200 inherits the full 3A090 three-layer mechanism; one carrying an L4 or a T4 does not. Screen the configured SKU, not the reseller.

QUOTE-ONLY FOR ACCELERATED SYSTEMS, and accelerator_ids is deliberately EMPTY - this is an honest negative. Servers Direct publishes transparent prices for BASE chassis (build-to-order towers, 1U/2U rackmounts and Xeon W workstations, listed as 'From now' discounted against a struck-through regular price), and markets itself explicitly for multi-GPU training and at-scale inference builds. But its 32-system catalogue as fetched on 2026-10-03 contained no GPU nodes and no priced accelerator SKUs; the GPU configurations are behind its 'Configure' / 'Get a Quote' path, which is where an H100/H200-class build would be priced. So the base-chassis prices it does publish must NOT be read as AI-node prices: they exclude GPUs, and for a 4- or 8-GPU node the accelerator is 80-95% of the cost. Recorded here as evidence of the reseller motion (site advertises 15% first-time-buyer discount, every system built to order and engineer-reviewed before ship) rather than as a price source. Its existence and pitch are verified from its own storefront. EXPORT CONTROL (2026-10-04): this record is a build-to-order reseller with NO named accelerator on the record - accelerator_ids is empty because the channel configures 'various OEM chassis to order' rather than carrying a fixed GPU SKU. Status recorded NULL rather than false or true: the control status is entirely determined by which GPU the customer configures into the box, and a server carrying an H100/H200/B200 inherits the full 3A090 three-layer mechanism while one carrying an L4 or a T4 does not. Do not treat the reseller itself as controlled or uncontrolled - screen the configured part. Sources: spl-eccn-3a090-ccl, spl-bis-2026-05-d5-guidance. LEAD TIME: null and it is an EXPLICIT NEGATIVE, checked 2026-10-04. This is a quote-only channel - no price and no lead time is published on any page, and AGENTS.md forbids inventing an unsourced number, so the field stays null with the reason recorded. The negative is itself a procurement fact: a channel that will not state a lead time is a channel where the lead time is set by whatever allocation you happen to have, and it is exactly the kind of number to ask for in writing before signing. Probe log and per-vendor detail in source spl-negative-no-oem-publishes-lead-time; the one figure published anywhere in this chain is AWS's 8-week MAXIMUM Capacity Block booking horizon, which is a commitment ceiling rather than a delivery time.