Best offer on record: Free rate-limited NIM hosted inference (NVIDIA Developer Program). 16 verified findings, checked 2026-10-22.
Offers
Free rate-limited NIM hosted inference (NVIDIA Developer Program) high confidence
Free API access to the NIM API Catalog (160+ models, incl. Nemotron 3 Ultra 550B, DeepSeek V4.x, GLM 5.3, Kimi K3, Llama 3.x) at build.nvidia.com, callable via OpenAI-compatible base URL https://integrate.api.nvidia.com/v1. Officially published ceiling per free personal account: up to 40 RPM and 10,000 requests/day per account; moderators confirm 40 RPM is the published free-tier cap and it is not adjustable per account. Governed by the NVIDIA API Trial Terms of Service (PDF last modified 2025-10-07): trial/prototyping purposes only, 'without use of the API Service or Generated Content in production'; NVIDIA may extend 'trial service credits' but must have a paid Subscription for production use. No credit card required; no per-token price. Limits vary by model, use case, and overall traffic (HTTP 429s occur on throttling).
Eligibility
Anyone with a free NVIDIA Developer Program account (no .edu email, no corporate email, no funding requirement). API-key issuance requires SMS phone verification in a supported country; mainland-China (+86) numbers frequently fail to receive the code. Not offered in sanctioned/embargoed regions: Russia, Belarus, Iran, Kazakhstan, Ukraine-adjacent set incl. Uzbekistan, Kyrgyzstan, Pakistan, Bangladesh plus 6 others (build.nvidia.com publishes an otp-unsupported-countries list; ~5.5% of developers worldwide excluded).
How to claim
1) Create an NVIDIA account at build.nvidia.com (Developer Program is free). 2) Verify phone number via SMS OTP (country must be supported). 3) Generate a personal API key in the API Keys section. 4) Call https://integrate.api.nvidia.com/v1 (OpenAI-compatible chat/completions) with model IDs copied from the catalog (only endpoints marked 'Free Endpoint'). Support: forums.developer.nvidia.com NIM section.
Expiry
No expiry while the Developer Program and the offering remain live; individual free endpoints rotate (models get deprecated and replaced by newer versions, routine catalog rotation). Free access lasts for the 'duration of program membership'.
Gotchas
Trial, not a permanent free tier: production use (serving real end users, business transactions) requires a paid NVIDIA AI Enterprise license. Inputs/outputs are recorded and used to improve NVIDIA products incl. AI models (Trial Terms 3.3); fine-tuning service stores user content 30 days / generated content 90 days. Arbitration + class-action waiver (30-day opt-out window). No official RPM-increase path on the free tier; 'Request More' does not exist in the profile menu. Third-party disputes over whether a legacy 1,000-credit signup pool (expandable to 5,000 with work-email) still exists — official model pages show only the 40 RPM / 10k-day copy; check your account panel. New personal keys can 404 on some chat calls until endpoints provision.
Free self-hosted NIM license (downloadable NIM containers, up to 16 GPUs) medium confidence
Developer Program members may download NIM microservice containers from NGC and self-host them for research, application development, and experimentation on up to 16 GPUs on any infrastructure (cloud, data center, workstation) at no charge from NVIDIA — your only cost is your own GPU infrastructure. Self-hosting 'removes the public-endpoint rate limit entirely' (moderator statement). Additional GPUs 'upon request'.
Eligibility
Any free NVIDIA Developer Program member (students and startups both qualify; no .edu or funding requirements, no CC needed). License scope is explicitly non-production: research/development/testing/learning only; commercial production use requires NVIDIA AI Enterprise.
How to claim
1) Join the free Developer Program. 2) Browse NGC catalog / build.nvidia.com model cards, pull NIM container (pull command + Helm chart shown under each model's Deploy tab). 3) Deploy on your own NVIDIA GPU(s), up to 16 for free dev/test.
Expiry
Tied to Developer Program membership; no published end date. Terms change with the NVIDIA Software License Agreement.
Gotchas
16-GPU cap: commercial/production deployments are a license violation without AI Enterprise. You still pay for GPUs (cloud or hardware). This offer is the 2024-era framing from the official NIM FAQ; current self-host pricing is also bundled into the paid AI Enterprise product (from ~$4,500/GPU/yr).
NVIDIA AI Enterprise 90-day free trial license (self-hosted, production-grade) medium confidence
One-time 90-day evaluation license for the full NVIDIA AI Enterprise suite (NIM microservices, NeMo, Omniverse, K8s tooling; NVIDIA-hosted trial also mentions a free trial of NIM endpoints). Each evaluation license includes 128 concurrent users. Trial clock starts when the activation key is redeemed in the licensing portal (nvid.nvidia.com/dashboard). Renewal: no auto-renewal; a second 90-day eval license can be requested via a form, and longer POCs are referred to NVIDIA Partner Network partners.
Eligibility
Organizations with corporate/business email — personal emails (Google/Yahoo etc.) are rejected on application. Compatible infrastructure required (NVIDIA Certified Systems / product support matrix); individuals without NVIDIA hardware are steered to NVIDIA sales. No credit card or payment info visible in the registration flow; it is a license, not a bill.
How to claim
1) Go to nvidia.com/en-us/data-center/products/ai-enterprise/get-started/ → 'Request Free 90-Day License' (or the Test Drive redirect URL). 2) Register with a corporate email at enterpriseproductregistration.nvidia.com. 3) Wait ~24h for approval email with license-portal access. 4) Redeem the Product Activation Key, download software from NGC, run a license server.
Expiry
Exactly 90 days from activation-key redemption; one-time per org, extension via a re-application or partner POC. Trial includes Omniverse but NOT Run:ai (separate contact needed).
Gotchas
Personal/freemail addresses are auto-denied — students without an org email effectively cannot claim it, so the 90-day trial is an org/startup offer, not a student offer. 'Failed to acquire license' message = trial over. Production use of the free Developer-Program NIM endpoints remains disallowed even while holding a trial — the trial is for self-hosted eval on your GPUs.
NVIDIA Inception startup program (free membership, cloud credits, no equity) high confidence
Free global virtual program for AI startups: no application fee, no membership fee, no equity taken; 'no deadlines or cohorts' — apply anytime, review ~2-4 weeks. Members get: free cloud credits from NVIDIA and partners (NVIDIA publishes NO fixed dollar figure on the page; third-party 2026 guides cite up to $100,000 in NVIDIA DGX Cloud credits for H100 capacity and up to $100,000 AWS credits via AWS Activate, plus free NIM API endpoints for prototyping), free NVIDIA Deep Learning Institute self-paced courses + discounted workshops, preferred pricing on select NVIDIA hardware/software (rebate off supplier price), full Developer Forum access, VC introductions via Inception Capital Connect, co-marketing badges/events. A 30% DGX Cloud discount tier reportedly requires a 4-node / $75,000 minimum spend (funded teams).
Eligibility
Company must be: officially incorporated, less than 10 years old, with a clear use case for NVIDIA technology (AI/ML/CV/robotics/physical AI/simulation/HPCC), actively operated by at least one founder, with a working website. No revenue, no VC funding, no US incorporation required. Excluded: consulting/outsourced-dev firms, crypto-associated companies, cloud service providers, resellers/distributors, public companies. Students/individuals cannot apply (incorporated company required).
How to claim
1) Go to nvidia.com/en-us/startups/ → Apply. 2) Sign in with an NVIDIA Developer account, fill company profile (name, site, founding date, team, focus area). 3) Describe the product in 3-4 sentences with explicit AI/GPU angle. 4) Submit; after review a partner manager activates benefits (credits, NIM, training) via intro call. Contact: inceptionprogram@nvidia.com.
Expiry
Membership is open-ended (no published end date; benefits 'for the duration of membership'). Credit expiry: NVIDIA does not publish on the official page — partner credit windows vary (AWS Activate typically 1-2 years).
Gotchas
No published dollar figure exists on NVIDIA's page for cloud credits — the $100K DGX Cloud and $100K AWS numbers come from third-party 2026 guides, so verify with your account manager before budgeting. Credits burn out while the ongoing GPU bill continues. No guaranteed access to scarce GPUs. Preferred pricing is a rebate, not a guarantee of supply.
NVIDIA Innovation Lab / DGX Cloud Innovation Lab for Inception startups medium confidence
Select NVIDIA Inception members can apply for the NVIDIA Innovation Lab (a.k.a. DGX Cloud Innovation Lab): ~2 months of hands-on access to NVIDIA DGX Cloud infrastructure (H100-class GPUs), integrated AI software, pre-optimized environments, and direct NVIDIA expert support/benchmarking guidance. NVIDIA's own page claims member results 'up to 17x faster performance and 90% shorter optimization cycles'. Tier-based: entry Inception gets portal access; Premier/selected members get the Lab with 'larger allocation'.
Eligibility
Must already be an accepted NVIDIA Inception member (which itself requires: incorporated, <10 years old, working website, NVIDIA use case, one active founder; excludes crypto, resellers, cloud providers, public companies). 'Purpose-built for early-stage' members; selection is invitation/application-based, not automatic for all Inception members.
How to claim
1) Apply to and be accepted into NVIDIA Inception (nvidia.com/en-us/startups). 2) Once in the member portal, apply for/complete the DGX Cloud Innovation Lab (selected members are invited; the FAQ says 'select members are invited to apply'). 3) Work with the assigned NVIDIA expert during the 2-month window.
Expiry
The Lab itself is a fixed ~2-month engagement per application; the parent Inception membership is open-ended. No published repeat-access policy.
Gotchas
It is a compute grant, not a perpetual free tier — after 2 months the same 30% discount / paid DGX Cloud pricing applies. Availability is discretionary ('select members'), so acceptance into Inception does not guarantee the Lab. No equity taken by NVIDIA either way.
Free hosted NIM endpoints covered by the API Trial Terms (data/usage caveats for everyone, incl. students) high confidence
The formal NVIDIA API Trial Terms of Service (PDF, last modified 2025-10-07, read in full) govern all free build.nvidia.com inference: access is 'for limited trial purposes only and without use of the API Service or Generated Content in production' (§1.2). NVIDIA 'may extend trial service credits' (amounts per API service; §1.4) — after credits are exhausted, production requires a separate paid Subscription. §3.3: NVIDIA collects session metrics, error/execution logs, feedback, and User/Generated Content 'to improve NVIDIA products and services, including AI models', and usage is logged for security/fraud and shared with third-party providers. §2.4: fine-tuning service stores user content 30 days and generated content 90 days unless a subscription is purchased. §11.2: NVIDIA may terminate your use 'at any time without notice'.
Eligibility
Applies to every free-tier user (students, individual devs, startups alike); requires an adult of legal age in the country of use, Developer Program account + phone-verified API key, and a supported country (sanctioned-region exclusions apply). No .edu email, no CC, no funding cap.
How to claim
Nothing to claim — it is the legal layer under the free tier above. Read the Trial Terms PDF linked from every build.nvidia.com model card ('The trial service is governed by the NVIDIA API Trial Terms of Service').
Expiry
No expiry of the terms themselves; each trial/credit grant, model endpoint, or service can be terminated by NVIDIA at any time (§11.2, §1.3 pre-release clauses). Arbitration/class-waiver opt-out must be sent within 30 days of starting use to be valid (§14.4).
Gotchas
Do not send PII, financial/health/government data or PCI data (§4.3). Do not use free endpoints to build a product that 'compete[s] with the API Service' (§4.12). No sublicensing or redistribution of outputs (§4.2). Because content can be used to improve NVIDIA models, treat the free tier as unsuitable for confidential workloads. 'Free' ≠ perpetual: endpoints and rate limits can change or be sunset at NVIDIA's discretion with no liability.
NVIDIA Inception Program (startups) — free membership & core benefits high confidence
Free, no-equity program: free self-paced NVIDIA Deep Learning Institute courses, discounted expert-led workshops, free cloud credits from NVIDIA and partners, preferred pricing (rebate on select hardware/software), NVIDIA Developer Forums access, Inception Capital Connect (VC exposure), badges/co-marketing. No application fees, deadlines, or cohorts. Note: the task's entry URL nvidia.com/en-us/ai-catalog/inception/ now 404s; the live official page is nvidia.com/en-us/startups/ (verified today).
Eligibility
Company must be officially incorporated, less than 10 years old, actively operated by at least one founder, with a clear NVIDIA technology use case; any funding stage including bootstrapped; global, no stated region limit. Excluded: consulting/outsourced-dev firms, crypto-related companies, cloud service providers, resellers/distributors, public companies. NOT a .edu program — no student individual track, a legal entity is required.
How to claim
1) Apply at nvidia.com/en-us/startups/ (portal: programs.nvidia.com/phoenix/application). 2) Company is reviewed; keep profile/product details current in the portal to remain eligible. 3) Request benefits (cloud credits, partner offers, courses) from the Inception benefits catalog; partners approve individually. No credit card required for membership.
Expiry
No stated expiry for membership; credits have partner-defined validity (e.g., Yotta 24 months, GCP 2 years).
Gotchas
'Free cloud credits' is a per-benefit request, not a fixed grant — amounts depend on which partner benefit you request. Preferred pricing is a rebate via suppliers, not a discount NVIDIA can guarantee. Membership can lapse if the portal profile goes stale.
NVIDIA DGX Cloud credits — up to $100,000 for H100 capacity (Inception members) medium confidence
Announced at GTC 2025 (March 2025 press release 'NVIDIA Expands Cloud Benefits for Startups'): NVIDIA offers qualified Inception members up to $100,000 in NVIDIA DGX Cloud credits for dedicated NVIDIA H100 GPU capacity, plus expert office hours. The figure still appears on NVIDIA's official Cloud Partners page as of today (2026-10-22), but the Inception main page does not list the benefit and 'qualified' criteria are unpublished; current availability not reconfirmable today.
Eligibility
Inception members (see base program: incorporated, <10 yrs old, any funding stage). 'Qualified members approved by NVIDIA after requesting the benefit in the portal' — qualification criteria are not public. No student path.
How to claim
As an Inception member, request the DGX Cloud credit benefit in the Inception benefits catalog; NVIDIA reviews and may approve up to $100,000. No CC required by NVIDIA itself.
Expiry
Not published.
Gotchas
Gated request, not automatic; no dollar amount promised at membership; criteria unpublished; if NVIDIA has wound this down, it was quietly — the Cloud Partners page still advertises it (stale-risk).
NVIDIA Cloud Partner free credits via Inception — Nebius up to $150K (+$10K inference), Yotta up to $50K, plus Lambda/Scaleway/Lintasarta medium confidence
GTC 2025 (April 2025) expansion: Nebius 'AI Lift' — up to $150,000 in Nebius AI Cloud credits (NVIDIA GPUs) plus $10,000 specifically for inference workloads; first 3 months of the program also gave priority Blackwell access. Yotta — up to $50,000 in credits in three tiers ($10K/$15K/$25K), valid 24 months (confirmed on Yotta's official program page today). Lambda and Scaleway offer credits; NVIDIA said these partner/ISV offers total ~$300,000 in credits and discounts combined. Each partner sets its own amount/terms and approves the request.
Eligibility
NVIDIA Inception or NVIDIA Connect members who request the specific partner benefit in the Inception portal and are approved by that partner. Global; partner-specific terms apply (Yotta's $15K/$25K tiers need more detail about your NVIDIA-platform usage).
How to claim
1) Join Inception. 2) In the benefits catalog, request the desired partner (Nebius/Yotta/etc.). 3) Partner reviews and issues credits on its platform; Yotta credits valid 24 months.
Expiry
Yotta: 24 months from grant; others partner-defined.
Gotchas
Amounts are partner-defined, not NVIDIA's; partner approvals can be slow. Yotta tiers require describing how you use NVIDIA SDKs/platform. The $300K 'total' NVIDIA headline is a marketing aggregate, not stackable credits.
AWS Activate via Inception (NVIDIA-issued Org ID) — Portfolio up to $200,000; self-funded Founders up to $5,000; invite-only AI tier $200,000+ high confidence
Official AWS Activate page (verified today): Activate Portfolio = up to $200,000 in credits (pre-Series B, requires Org ID from an Activate provider; NVIDIA is an approved provider via Inception). Activate Founders (self-funded/bootstrapped) = start with $1,000, up to $5,000 total, apply directly without Org ID. New 'AWS Credits for AI Startups' = $200,000+, invite-only for scaling startups via account manager. A new 'Kiro Startup Credits' track also appears. Credits are redeemable on third-party models on Amazon Bedrock; 'AWS Account on Paid Tier Plan' is required. NVIDIA developer forums (2025–2026) show members previously receiving $10K/$25K tiers and up to $75K–$100K Scale-tier approvals, plus persistent Org-ID handoff issues between Inception and AWS.
Eligibility
Founded within the last 10 years; Pre-Series B (Portfolio); most recent funding round rules apply on AWS's side; bootstrapped founders use the Founders tier directly; for Inception-referred applicants: Inception membership + NVIDIA-confirmed AWS benefit + NVIDIA Org ID. AWS paid-tier account (credit card on file) required.
How to claim
Inception route: 1) Join Inception, request the AWS benefit in the portal. 2) NVIDIA confirms eligibility and issues an AWS Activate Org ID. 3) Apply on AWS Activate (startups.aws.com) using the Org ID; AWS reviews (5–10 business days per AWS; community reports months). Direct route (self-funded): apply directly to Activate Founders, no Org ID. No .edu/student track (individuals just need an AWS account; students have AWS Free Tier separately).
Expiry
Activate credits validity is set by AWS promotional terms (typically 12 months from grant) — not restated on the page today.
Gotchas
Official tiers were restructured — the older member-reported $10K/$25K/Scale-up-to-$100K offers no longer appear on AWS's current page; the handoff (NVIDIA Org ID → AWS) has documented 3–15 month delays and auto-rejections ('already received credits at same or greater amount', 'provider has not confirmed affiliation'). Paid-tier account = CC required. Bedrock third-party model usage consumes credits.
Google Cloud for Startups Cloud Program — up to $350,000 credits over 2 years (AI track) high confidence
Official GCP page (verified today): 'Up to $350,000 in Cloud credits over two years', plus credits for frontier models (Gemini/Gemma), technical/product training. Tier structure (footnotes on the page): AI startups — Year 1: up to $250,000 at 100%; Year 2: up to an additional $100,000 at 20%. Other startups — Year 1: up to $100,000 at 100%; Year 2: up to an additional $100,000 at 20%. 'Most tiers require verifiable Pre-Seed to Series A funding.' Third-party models are billed directly, not covered by credits. This program is a stated Inception partner benefit (NVIDIA↔GCP collab since April 2024 offered the up-to-$350K path through Inception).
Eligibility
Incorporated startup, verifiable Pre-Seed to Series A funding for most tiers; AI track requires the product to be AI-focused; apply at cloud.google.com/startup/apply. Inception members get an accelerated onboarding path (partner deal). No .edu requirement; not open to individuals/students. Card policy not restated on the page today (GCP account required; standard GCP trials normally take a card on file).
How to claim
1) Apply at cloud.google.com/startups (Apply to the program). 2) Verify funding stage / company details. 3) If Inception member, use the partner fast-track. 4) Draw down: AI — up to $250K year 1; +up to $100K at 20% (i.e., up to $20K effective in year 2 per footnote wording). 5) Credits cover Google-built models; third-party model inference is billed separately.
Expiry
Credits usable over 2 years as structured above.
Gotchas
'Year 2: 20% up to an additional $100,000' means only 20% of year-2 usage up to $100K is credited, not the full amount — real max over 2 years is less than a naive $350K + $100K read. Third-party (e.g., non-Google) model API calls are excluded. Pre-Seed–Series A gate excludes bootstrapped/late-stage companies (unlike AWS's pre-Series B).
Google Cloud free trial + always-free tier — $300 / new customers (open to students & individuals) high confidence
Official page (verified today): 'New customers get $300 in free credit to try Google Cloud products and build a proof of concept. You won't be charged until you activate your full paid account.' Plus 20+ always-free products with monthly limits ('$300 in free credits and 20+ free products'), free trial instances for GKE, Cloud SQL, Spanner, AlloyDB, Bigtable, plus Gemini Enterprise Plus free trial. This is the individual/student-reachable free GPU/cloud front of the ecosystem (GCP A-series GPU instances can be run on the $300 trial).
Eligibility
New GCP customers; open to anyone with a Google account (students with .edu or personal both qualify); no funding/age requirements; global. Standard signup takes a credit/debit card on file (not restated in today's page text — historical standard).
How to claim
Sign up at console.cloud.google.com/freetrial; $300 credit is valid for 90 days under standard terms; always-free quotas apply regardless. Use on A3/A100/A100M GPU instances or Vertex AI for ML training jobs.
Trial is a 90-day time-boxed trial, not perpetual free compute; if the paid account activates after trial expiry, standard pay-as-you-go applies; card required (verify current signup). GPU availability for A100/A3 is region-limited.
Microsoft for Startups — 'up to $150,000 in credits' on Azure + free AI tools & GitHub credits medium confidence
Official hub (verified today): 'Access cutting-edge AI models, developer tools, and up to $150,000 in credits' — a large 2025/2026 expansion vs the legacy $1,200/year Azure + $1,200 GitHub framing. Credits apply to eligible Azure services on a sponsored subscription; Microsoft Learn (today) documents credit-aware use across GitHub (Copilot, Actions, billing connected via the Azure subscription), Microsoft Foundry models, Fireworks models, and AKS (AKS usage on a sponsored subscription is auto-covered). Tier details and eligibility are gated behind the application flow, not published on the marketing page.
Eligibility
Incorporated startups; Microsoft's own 'Microsoft for Startups' application decides tier/amount; Inception members can apply independently (no NVIDIA Org ID involved). No .edu/student track — it is a startup program. Exact funding/age caps not stated on the pages verified today.
How to claim
1) Apply via Microsoft for Startups (www.microsoft.com/en-us/startups → apply/claim benefits). 2) Claim 'Startup credits' into a sponsored Azure subscription. 3) Optionally connect that subscription as GitHub's payment method so GitHub Copilot/Actions metered usage bills against the credits. No CC required for the sponsored credits themselves (standard Azure account still needed).
Expiry
Not published on verified pages; Microsoft historically used 12-month renewal windows for Azure startup credits.
Gotchas
'Up to $150,000' is a marketing ceiling — the bulk of recipients get far less; published tier table not found on marketing pages. Partner/community models on Foundry are billed separately from Microsoft-sold models. Legacy $1,200 figure no longer appears — check the application flow for what you will actually be offered.
NVIDIA Inception Innovation Lab — 60 days of hands-on NVIDIA GPU access (selected members) medium confidence
Inception benefit: 60 days of hands-on NVIDIA GPU access on a self-serve platform with pre-optimized environments and expert support, run in cohorts; NVIDIA provides the cloud instance (no dollar value published). Reported in the current (2026) Inception benefits summary; it is a selected-member benefit, not automatic.
Eligibility
Selected NVIDIA Inception members (cohort-based); same base Inception eligibility (incorporated, <10 yrs, NVIDIA use case).
How to claim
Join Inception, then be selected into an Innovation Lab cohort; request via the benefits catalog.
Expiry
60 days per cohort.
Gotchas
Cohort selection — approval not guaranteed; time-boxed 60 days; no published dollar value; availability fluctuates.
build.nvidia.com — free NIM inference for open models incl. Nemotron (open access, incl. students) medium confidence
Official build.nvidia.com (verified today): 'Free inference with leading models' — free API endpoints for Nemotron models (and other open models, e.g., DeepSeek) plus pre-built agentic capabilities and blueprints. Open sign-up (no startup status or .edu stated as a requirement), so this is the individual/student-reachable 'free ML compute' front, but it is inference API, not GPU training or notebooks. Community reports a default RPM cap around 40; Inception members can request higher RPM for research workloads (forum, Oct 2026).
Eligibility
Any NVIDIA developer account; Inception members may request raised rate limits. No funding/institution gate documented. No CC stated for inference use.
How to claim
1) Sign in at build.nvidia.com. 2) Use the model's API endpoint / key to run inference free within rate limits. 3) Inception members: request higher RPM via the Inception portal / developer forums for research benchmarks.
Expiry
No stated expiry; free tier subject to change.
Gotchas
Inference-only — cannot train or host notebooks here; rate-limited; free-tier availability of specific models can change; heavier workloads require paid or partner-cloud (Inception credit) routes.
NVIDIA education angle — free self-paced DLI courses; no dedicated student GPU-compute program found medium confidence
NVIDIA's student/education-free offer is educational, not compute: free self-paced Deep Learning Institute courses (course codes distributed via Inception membership; public DLI catalog is also free), discounted instructor-led workshops, free CUDA Toolkit/NGC developer downloads. No NVIDIA 'education license' or .edu GPU quota program (e.g., Colab-style notebooks) was found in the Inception or NVIDIA education pages as of today; student GPU compute in this ecosystem effectively comes through the partner free trials (GCP $300/90-day + free A-series eligibility, AWS Free Tier, Azure) or via starting a company and joining Inception.
Eligibility
Individuals/students: free DLI courses, build.nvidia.com inference, partner free trials (no .edu email specifically required by NVIDIA; .edu mainly matters at third parties like Google's student pricing elsewhere). Startups: Inception (incorporated entity).
How to claim
Students: enroll in DLI courses / use build.nvidia.com free inference / claim GCP $300 trial and AWS Free Tier on a personal account. Startups: apply to Inception for the partner-credit stack above.
Expiry
N/A (courses free; trials per partner terms).
Gotchas
Do not conflate 'free trials' (time-boxed, card on file, pay-as-you-go after) with 'free' — only DLI, build.nvidia.com inference (rate-limited), and partner credits are genuinely free.
Education/student-specific offers: as of 2026-10-22 I could NOT verify a dedicated NVIDIA education license or student program for inference/AI Enterprise (NVIDIA education-related URLs www.nvidia.com/en-us/education/ and /about-nvidia/education/ return 404; my web-search budget for this session was exhausted, so I could not confirm whether a GTC-era 'AI for Students' pilot at partner universities still exists). What is verifiably available to students is: the free Developer Program NIM tier (finding 1) and the free 16-GPU self-hosted NIM license (finding 2) — both open to individual students with no .edu requirement, plus free DLI courses if their startup is in Inception. A dedicated student free-offer line could not be confirmed and should not be cited. Sunset/changes 2025-2026: (a) The legacy '1,000 signup credits, expandable to 5,000 with work-email verification' mechanism reported on the NIM forum in 2024-2025 has effectively been replaced by the published 40 RPM + 10,000 requests/day rate-limit model — official model pages show no credit counter; 2026-08 forum threads still use the old framing, NVIDIA has not officially confirmed the credits pool was removed (disputed; check account panel). (b) The 2024 official NIM FAQ's framing ('160+ models', '90-day trial' linked off the dev-program page) is superseded by the 2026 catalog rotation (Nemotron 3.x/DeepSeek V4.x/GLM 5.3 generation) and by consolidation of Omniverse terms into 'NVIDIA AI Enterprise' (AI Product Terms last modified 2026-04-15). (c) DGX Cloud's old 'free trial' web page (nvidia.com/.../free-trial-nvidia-ai-enterprise) now 301-redirects to the corporate-email enterprise registration portal, meaning the self-serve free trial path has been tightened to organizations with corporate emails. (d) Rate limits are explicitly non-adjustable on the free tier; moderators direct all RPM-increase requests to paid AI Enterprise channels (2026 forum replies). Inception publishes no dollar figure for its cloud credits; widely-cited 'up to $100,000 DGX Cloud' number is third-party-sourced only. Regional note: the free tier is unavailable in Russia, Belarus, Iran, Kazakhstan, Uzbekistan, Kyrgyzstan, Pakistan, Bangladesh and 6 other countries (NVIDIA's own otp-unsupported-countries config, read 2026-09-26).
2025-2026 changes/sunsets: (1) The task's entry URL https://www.nvidia.com/en-us/ai-catalog/inception/ now returns 404 — Inception's live official page moved to https://www.nvidia.com/en-us/startups/ (verified today). (2) AWS Activate restructured: the member-reported Inception-ecosystem tiers ($10K base / $25K Grow / up to $100K Scale, per NVIDIA developer-forum threads 2024-2025) no longer appear on AWS's current official page; the official structure is now Activate Founders (self-funded, up to $5,000, start $1,000), Activate Portfolio (pre-Series B, up to $200,000, NVIDIA Org ID via Inception), and invite-only 'AWS Credits for AI Startups' ($200,000+), plus a new 'Kiro Startup Credits' track. Documented 3-15 month handoff delays (Org ID issues) persist in late-2025/2026 forum threads. (3) GCP for Startups is still up to $350K, now formally structured over 2 years (Year-1 100% / Year-2 20% footnotes, up to $250K Y1 AI / $100K Y1 others) with verifiable Pre-Seed to Series A funding required for most tiers — the older flat 'up to $200,000' marketing was effectively superseded by the $350K AI headline. The $300/90-day GCP free trial is unchanged. (4) Microsoft for Startups headline credits expanded from the legacy $1,200/yr Azure + $1,200 GitHub to 'up to $150,000 in credits' with Foundry/GitHub credit-aware usage (current official hub). (5) New since 2025: NVIDIA DGX Cloud H100 credit cap (up to $100,000, GTC 2025) and Nebius AI Lift (up to $150K + $10K inference, GTC April 2025). Nothing on the GCP $300 trial, GCP always-free tier, or build.nvidia.com free inference was sunset; Microsoft and AWS program details not published beyond the levels quoted. No confirmed full sunset of any offer in this stack as of 2026-10-22, though the $100K DGX Cloud offer's current operational availability is unconfirmable (still advertised on NVIDIA's Cloud Partners page but absent from the main Inception page).