AWS EC2 Spot for GPU: the 90% ceiling is real but the per-SKU rate is not published in one place

NO SINGLE SPOT price_usd, deliberately, and the basis is stated as a CEILING not a rate. AWS states Spot is 'available at a discount of up to 90% off compared to On-Demand pricing' and that 'Spot Instance prices are set by Amazon EC2 and adjust gradually based on long-term trends in supply and demand'. The page's own table 'displays the Spot price for each region and instance type (updated every 5 minutes)' but states 'The prices displayed below are the lowest prices per instance type in the region' - so even the rendered grid is a FLOOR per instance type, not a per-AZ rate, and the per-AZ numbers live behind the Spot Instance pricing history page. Spot price history is only reachable through the describe-spot-price-history API. The concrete On-Demand anchors these discounts apply to, read out of the EC2 Price List API us-east-1 file with publicationDate 2026-09-25: p5.48xlarge (8x H100 SXM) $55.04/instance-hour; p5.4xlarge (1x H100 SXM) $6.88; p5en.48xlarge (8x H200) $63.296; p4de.24xlarge (8x A100 80GB) $27.44705; p4d.24xlarge (8x A100) $21.957642; g6e.48xlarge (8x L40S) $30.13118; g6e.xlarge (1x L40S) $1.861; g6.48xlarge (8x L4) $13.3504; g4dn.12xlarge (4x T4) $3.912. PER-GPU-HOUR, derived: H100 SXM $6.88, H200 $7.912, A100 80GB $3.4309, L40S $3.7664, L4 $1.6688.

lead_time_weeks is null because the quantity is structurally not applicable: on-demand cloud capacity starts on request, so there is no order-to-delivery lead time. The constraint is the price per GPU-hour and the availability snapshot, not an interval. The one caveat worth keeping: scarce tiers gate capacity behind a reservation or Capacity Block commitment (AWS caps the booking horizon at 8 weeks, source spl-aws-capacity-block-booking-horizon), and that is a commitment window, not a lead time.

EAR 742.6(a)(6)(iii)(A) + EAR 744.23/744.6 (Entity, Unverified, MEU list screening of the end user) + the advanced-computing IaaS obligations of 91 FR 1684 (consignee KYC, and management/limitation of IaaS access to the hardware)

LAYER 1, EXPORT LICENCE: under the Jan 2026 BIS rule 91 FR 1684 (effective 2026-01-15) the license review policy for advanced-computing commodities with TPP < 21,000 AND total DRAM bandwidth < 6,500 GB/s - the NVIDIA H200 and AMD MI325X are the rule's named examples - moved from PRESUMPTION OF DENIAL to CASE-BY-CASE REVIEW, but only for exports FROM THE UNITED STATES to end users located in China or Macau, and only against six exporter certifications filed via SNAP-R before shipment: (1) items below the thresholds with the count of units shipped in the US; (2) sufficient US supply, judged 'taking into account normal lead times'; (3) no diversion of global foundry capacity away from US end users; (4) aggregate China+Macau shipments <= 50% of total US end-use shipments of that product; (5) no end-user or use prohibition; (6) the consignee runs rigorous KYC. Every shipment must then pass a qualified third-party testing lab - US-headquartered, not controlled by a D:5/Macau entity, no ownership stake in any party, qualification revocable by BIS at any time - verifying TPP, total DRAM bandwidth, interconnect bandwidth and copackaged DRAM capacity. Reexports and in-country transfers of the same chips to China/Macau REMAIN PRESUMPTION OF DENIAL. LAYER 2, END-USER AND PARENT SCREENING: per BIS guidance of 2026-05-31 a licence is required where the consignee is HEADQUARTERED in Country Group D:5 (15 CFR part 740 supplement no. 1) or Macau OR whose ULTIMATE PARENT COMPANY is so headquartered - even when the consignee itself sits outside D:5/Macau. The requirement sits at EAR 742.6(a)(6)(iii)(A) for destinations outside the US, with an exception available only under the license exception at 740.2(a)(9)(ii). The destination country alone therefore never answers the question; screen the parent. LAYER 3, REMAINING LIMITS: exports to D:5/Macau-headquartered OR -parented entities stay PRESUMPTION OF DENIAL wherever the buyer sits, and the AI Diffusion Rule that would have set a country-tier structure on who could lawfully buy what was RESCINDED on 2025-05-13 and is not being enforced, so any recorded status rests on the instruments named here rather than on tiering.

THREE THINGS THAT MAKE 'UP TO 90% OFF' UNUSABLE AS A BUDGET NUMBER. (1) It is a ceiling across ALL EC2 SKUs, not an H100 figure; nobody has ever paid 90% off for p5. (2) The published grid is the LOWEST price per instance type in the region, so budgeting off it understates what you will actually pay in a specific AZ, and GPUs are the SKUs where per-AZ spread is widest because capacity is lumpy. (3) There is no single place on AWS's site where a per-SKU spot rate is stated; the only authoritative source is the describe-spot-price-history API or the Bid Advisor page. The interruption reality from AWS's own documentation: Spot instances are reclaimed on a two-minute warning, and p5/p5en are the SKUs with the deepest demand pools and therefore the worst interruption profile - which is a structural argument, not a measured one. The commitment-discount exclusion is the rule that breaks naive cost models: on Google Cloud the pricing page states explicitly that Spot VMs and preemptible VMs cannot receive CUDs or SUDs. Never model a spot instance as also earning a commitment discount. Related existing records: aws-ec2-spot-gpu-general for the structural picture, aws-ec2-p5-h100 and aws-ec2-p5en-h200 for the on-demand anchors. EXPORT CONTROL (2026-10-04, mechanism-level; sources spl-bis-2026-05-d5-guidance, spl-bis-2026-01-advanced-computing-case-by-case): renting capacity does NOT remove the export-control question, it relocates it. Providing remote access to controlled computing capability can constitute a transfer under the EAR, so the provider must screen the end user; and separately, under the 91 FR 1684 licence regime the consignee must run KYC procedures to prevent unauthorized remote access and must manage and limit IaaS access to the hardware - obligations pushed onto the customer, who must be told to implement them. For a customer headquartered or parented in Country Group D:5 or Macau the 742.6(a)(6)(iii)(A) end-user/parent test can bite even when the capacity is consumed in a permitted region. Recorded true because the underlying hardware is controlled and the provider carries the screening duty, not because access to a US region is itself restricted. LEAD TIME: null, and this is not a gap - for Spot it would be the wrong unit. Spot capacity is reclaimed at any time, so its defining characteristic is that there is no forward commitment available at all: the correct planning figure is the INTERRUPTION rate, not a delivery date, and Azure publishes the corresponding number directly ('Azure will evict Azure Spot Virtual Machines with 30-seconds notice') while GCP states Compute Engine 'might preempt Spot VMs at any time'. Anyone needing a guaranteed week-count should read the on-demand or Capacity Block rates instead; the 8-week maximum Capacity Block booking horizon is documented in source spl-aws-capacity-block-booking-horizon. Probe log in source spl-negative-no-oem-publishes-lead-time.