AWS credits for US startups — the Portfolio Sub-Program home base, every region, and the $150K stacked ceiling.

IWhy the US is the structural baseline for AWS Activate — and what that means in practice

AWS Activate, ACE (APN Customer Engagements), the Portfolio Sub-Program, Build for Startups, and the Bedrock POC funding mechanism were all designed around the US startup base. The published guides, the reviewer playbooks, and the Sub-Program-enrolled VC list all reflect a US-default. For US-incorporated startups, this means lower friction at every stage of the application — but it also means the smaller advantages that compound on top of the baseline are easy to miss.

The Portfolio Sub-Program — the mechanism by which a VC vouch generates a $100K Activate Portfolio credit allocation for a portfolio company — is operated primarily through US-headquartered fund LPs. The roster of directly-enrolled VCs is dominated by US tier-1 funds: Andreessen Horowitz (Menlo Park, NY, SF), Sequoia Capital US (Menlo Park, SF), Bessemer Venture Partners (Menlo Park, NY, Boston), Accel (Palo Alto, NY, London), Greylock (Menlo Park, SF), Founders Fund (SF), New Enterprise Associates (Menlo Park, DC, Chicago), Index Ventures (SF, NY, London, Geneva), Lightspeed Venture Partners (Menlo Park, NY, Bangalore, Tel Aviv), Insight Partners (NY), General Catalyst (Cambridge, SF, NY, Berlin via the La Famiglia merger), Tiger Global (NY), Coatue (NY, SF), Khosla Ventures (Menlo Park), and GV (Mountain View). The Sub-Program enrollment is fund-level — a Sequoia US partner vouch and a Sequoia India partner vouch route through different sub-program enrollments.

For a US-incorporated startup with backing from any fund on this list, the path to $100K Portfolio is direct: the fund's partner or platform team submits the Portfolio request to AWS through the Sub-Program portal, the request is auto-attributed to the fund's LP allocation, and the credits land in the founder's AWS account in 7–14 days. No partner-filed routing is needed; no ACE record is required; the Sub-Program flow is the canonical path.

For a US-incorporated startup without backing from a Sub-Program-enrolled fund, the partner-filed ACE route is the alternative. This is the same mechanism used for international startups — an APN Partner submits an ACE record on the startup's behalf, attesting to the project scope, projected consumption, and use case. Time-to-balance is 11–18 days, slightly longer than the direct Sub-Program flow but identical in ceiling ($100K for Portfolio; $5K–$25K for partner-filed Founders).

The accelerator-density advantage is the second US-specific structural factor. Roughly 90% of all Y Combinator-backed startups are US-incorporated; YC backing carries $5K standing AWS Activate credits and direct access to the Portfolio Sub-Program path. Techstars (US-based accelerators in Boulder, NY, Boston, Chicago, Austin, LA) carries similar mechanics. 500 Global, Antler (selective in US), On Deck, AI Grant, a16z Speedrun, Sequoia Arc, Y Combinator AI Startup School, Pear VC Accelerator, and Plug and Play Tech Center all generate accelerator signals that AWS reviewers in the US queue recognize. The typical US founder has at least one of these signals available, even if not always activated.

Wall-clock advantage in the US queue is the third factor. The AWS Americas review queue is the largest staffed queue at AWS, processing the majority of global Activate Portfolio applications. Median time-to-decision for partner-filed applications is 8–12 business days in the US queue, against 12–18 for EMEA and 14–22 for APAC. Sub-Program-direct applications resolve faster still — 5–10 business days median.

IIus-east-1, us-west-2, us-east-2, us-west-1, and the GovCloud separation

The US has the densest AWS region footprint of any country. Four commercial regions plus two GovCloud regions plus AWS Local Zones in 15+ metropolitan markets. The default selection patterns and the situations that override them are well-trodden — but the choices materially affect cost (us-east-1 is the cheapest commercial region for most service families), service availability (us-east-1 launches new services first), latency (sub-region differences matter at the millisecond level for some workloads), and compliance posture (GovCloud is a fundamentally separate environment).

us-east-1 (Northern Virginia) is AWS's oldest and largest region, operating since 2006. It has six Availability Zones (us-east-1a through us-east-1f), the most of any AWS region globally. Pricing in us-east-1 is the floor for most service families — EC2 on-demand pricing, S3 storage pricing, RDS pricing, and most managed-service pricing are set lowest in us-east-1 by default. New AWS services almost always launch in us-east-1 first, with rollout to other regions over the subsequent 3–18 months. For US-incorporated startups without a region-specific reason to choose otherwise, us-east-1 is the default region. Approximately 60% of CloudRoute-routed US startup engagements in 2025–2026 select us-east-1 as primary.

us-west-2 (Oregon) is the second-most-common primary region for US startups, particularly those headquartered in the Bay Area, Seattle, Portland, or Los Angeles. Three Availability Zones with a fourth historically operated. us-west-2 carries the full mainstream service catalog with the smallest lag behind us-east-1 (typically 1–2 release cycles). The region operates substantially on renewable energy — AWS reports the Oregon facilities as running on a regional power mix with significant hydroelectric and wind generation, which matters to startups with ESG-conscious enterprise customers or VC LPs. Bay Area to us-west-2 latency: 18–24ms; Seattle to us-west-2: 6–10ms; Los Angeles to us-west-2: 22–28ms.

us-east-2 (Ohio) opened in 2016 as the second East Coast region. Three Availability Zones. The strategic purpose of us-east-2 in most US startup architectures is multi-region resilience — pairing us-east-1 (primary) with us-east-2 (DR target) gives geographically separated infrastructure within the same compliance jurisdiction without crossing the country. Service availability in us-east-2 is roughly 90% of us-east-1 with a 3–9 month launch lag for newer services. Pricing is identical to us-east-1 for most services.

us-west-1 (Northern California) opened in 2009. Two Availability Zones (a historical configuration that has not been expanded). us-west-1 is the closest AWS region to San Francisco proper — Bay Area to us-west-1 latency is 4–8ms versus 18–24ms for us-west-2. The region is used by Bay Area startups with extreme latency sensitivity (real-time gaming, high-frequency-trading-adjacent workloads, real-time collaboration tools with sub-10ms latency budgets). The two-AZ configuration limits multi-AZ resilience patterns; most production workloads using us-west-1 pair it with us-west-2 for the third-AZ-equivalent.

AWS GovCloud (US) is a fundamentally separate AWS environment from the commercial regions. Two regions: GovCloud (US-East) in Ohio and GovCloud (US-West) in Oregon. GovCloud is operated by US-citizen personnel with screened access, runs under FedRAMP High authorization for the underlying infrastructure, and is logically and physically isolated from the commercial AWS network. Customer eligibility for GovCloud is restricted to US persons and US-located entities serving US federal, state, local, tribal, or territorial governments, or supporting federal contractors with ITAR/EAR-controlled workloads. Activate credits do not directly fund GovCloud consumption — the credit balance is tied to the commercial AWS account, not the GovCloud account. Startups operating both a commercial development environment and a GovCloud production environment typically use credits to subsidize commercial-side dev/staging while paying cash for GovCloud production.

AWS Local Zones are deployed in 15+ US metros (Boston, Chicago, Dallas, Denver, Houston, Las Vegas, Los Angeles, Miami, Minneapolis, New York, Philadelphia, Phoenix, Portland, San Diego, Seattle, Atlanta, Detroit, Kansas City, Nashville, Charlotte) and provide a subset of EC2, EBS, VPC, and load-balancer services with sub-10ms latency to the metro. Local Zones are extensions of a parent region (typically us-east-1, us-east-2, us-west-2, or us-west-1) — credit balances apply to Local Zone consumption identically to parent-region consumption.

IIIDelaware C-corp, Stripe Atlas, Mercury banking, and Activate Founders auto-issuance

The Delaware C-corp is the dominant incorporation pattern for VC-backed US startups — by some counts, 80%+ of YC-graduate companies and 90%+ of Series-A-stage companies are Delaware C-corps. This incorporation pattern interacts with AWS credit eligibility verification through a specific and well-trodden integration path: Stripe Atlas formation handoff, Mercury banking primary, and AWS Activate Founders auto-issuance through partnership listings.

AWS Activate Founders self-serve eligibility requires the applicant to operate a startup that is registered as a business entity, has a website, and has been founded within the past 10 years. The verification mechanism — when self-serve is filed — is light: the applicant enters company name, website, founding year, AWS account ID, and selects a self-attested funding stage. The auto-issuance flow is the path where AWS partners with formation and banking platforms to pre-verify the eligibility signals and apply credits directly upon account creation.

Stripe Atlas is the most common formation platform for US-incorporated startups in CloudRoute's routed pipeline. The Atlas product handles Delaware C-corp formation (Certificate of Incorporation filed in Delaware, registered agent designated, EIN obtained from the IRS, founder shares issued with 83(b) elections filed), bank account opening through partner banks, and post-formation handoffs to a network of partner services. AWS Activate is one of the standard handoff partners — when an Atlas-formed startup creates an AWS account using the Atlas-provided onboarding flow, the $5K Founders credit is typically auto-applied within 24–72 hours of AWS account creation. The mechanism is the Atlas dashboard's "AWS Activate" tile, which generates a partner-attested signup link that pre-fills the company info from the Atlas formation record.

Mercury Bank is the most common banking primary for early-stage US tech startups in 2026. Mercury maintains a "Mercury Perks" benefits program that includes AWS Activate credits as a default offer — typically $5K standing for new accounts, with documented escalation paths to higher tiers ($10K–$25K) for accounts meeting specified balance or transaction-volume thresholds. The Mercury-to-AWS handoff is similar to Atlas: a dashboard tile generates a partner-attested signup link, the link pre-fills company info, and the credit is auto-applied within 24–72 hours.

Brex (banking and corporate cards), Ramp (corporate cards and bill pay), and Carta (cap table management) all maintain similar AWS Activate handoff integrations. The credits accessed through these platforms are typically $5K standing — the same as direct self-serve — but the verification friction is reduced because the platform has already KYC-verified the company. CloudRoute observation: a Delaware C-corp formed through Stripe Atlas, banked at Mercury, with Brex corporate cards typically has three independent paths to $5K Activate credits (each platform offers it; the credits do not stack — only the first claim applies).

The Delaware C-corp pattern also affects credit eligibility for the higher tiers. The Portfolio Sub-Program flow assumes the portfolio company is a Delaware C-corp (or convertible to one), because the VC term sheets and post-money cap tables that the Sub-Program references are structured around Delaware C-corp share classes (Series Seed Preferred, Series A Preferred, etc.). A US LLC or partnership-structured startup applying for Portfolio Sub-Program credits with an institutional vouch typically receives a clarifying question from the reviewer about conversion-to-C-corp plans — not a denial, but an additional 3–5 day cycle.

The Delaware Division of Corporations' public search interface (icis.corp.delaware.gov) is the standard reference AWS reviewers use to verify Delaware C-corp status during the credit application review. The reviewer enters the company name, confirms the Delaware filing, and proceeds. For a recently-formed startup with the Delaware filing visible in the public search interface, this is 30 seconds of reviewer time. For a startup with formation pending or an in-flight conversion, the verification stalls until the public record updates — typically 3–7 days post-Delaware filing.

IVTypical credit pools for US startups — Founders, Portfolio, Build for Startups, Bedrock POC

The US credit envelope is the broadest and most consistently-stackable of any geography. The Series-A median stack — $100K Portfolio + $25K Build for Startups + $25K Bedrock POC = $150K combined — is the most reliably achievable composition globally. Below the Series-A tier, the partner-filed Founders track ($5K–$25K), the auto-issuance flows ($5K via Atlas/Mercury/Brex), and the accelerator standing offers (YC, Techstars, 500 Global) compose into smaller but still useful pools.

At the pre-seed / no-funding stage, a US-incorporated Delaware C-corp can stack: $5K from one of the auto-issuance flows (Atlas, Mercury, or Brex — typically only the first claim applies); $5K self-serve Activate Founders if not already claimed via auto-issuance; nothing further unless the company is accelerator-affiliated. Total typical pool: $5K, with validity of 12 months from issuance.

At the accelerator stage (YC, Techstars, 500 Global accepted but pre-Demo Day): $5K self-serve + accelerator-specific standing offers. YC carries $5K standing AWS Activate credits via the YC startup-services agreement; Techstars carries $1K standing with documented escalation paths; 500 Global carries $5K standing through similar partner integrations. These standing offers also serve as a credibility signal that lifts subsequent partner-filed Founders applications toward the upper end of the $5K–$25K range. Total typical pool at the accelerator stage without partner-filed routing: $5K–$10K.

At the seed stage with institutional backing from a Sub-Program-enrolled VC (any tier-1 US fund), the Portfolio path opens directly. The fund's platform team submits the Portfolio request through the Sub-Program portal; the credits land at the $100K Portfolio ceiling in 7–14 days. For seed-stage companies the typical landing is $50K–$80K (reviewers calibrate to projected consumption, and seed-stage projected consumption is typically lower than Series-A); for traction-positive seed companies the full $100K is achievable. Adding Build for Startups (+$25K) at this stage requires a distinct workload scope, which most seed-stage companies have once they identify the second product line or the SOC 2 compliance build-out.

At the Series-A stage, the full stack is reliably achievable: $100K Portfolio (Sub-Program-direct via the VC platform team, or partner-filed via ACE for non-Sub-Program-VC-backed companies) + $25K Build for Startups (distinct workload, typically scoped around SOC 2 prep, security tooling deployment, or a second-product-line build) + $25K Bedrock POC (Bedrock-earmarked, scoped around a customer-facing AI feature with documented evaluation methodology) = $150K combined. The Series-A median in CloudRoute's 2025–2026 routed pipeline is exactly $150K, achieved by approximately 70% of US Series-A applications.

Above $150K, the additional pools require either competitive admission or specific scope. The AWS Generative AI Accelerator is a competitive cohort program — typical accepted-cohort credits cluster around a $300K median with a $1M ceiling for the largest-funded participants. Acceptance to the Generative AI Accelerator is competitive (single-digit-percent acceptance rate for the inaugural cohorts; equalized since 2024 but US applicants remain the plurality of accepted companies due to AI-native company density in the SF and NY ecosystems). The Migration Acceleration Program (MAP) funds 25–50% of qualified migration costs for larger workloads — typically $200K–$500K+ for mid-market Series-B-and-beyond migrations off competing clouds.

A common confusion in the US credit math: the Bedrock POC pool ($10K–$50K) and the AWS Generative AI Accelerator ($300K median) are different programs. Bedrock POC is partner-filed via ACE with a defined POC scope (use case, model selection, evaluation methodology, projected spend, 60-day window typical). The Generative AI Accelerator is a competitive program with cohort-based admission, longer engagement windows (typically 10 weeks of structured programming plus the credit allocation), and dedicated AWS solutions architect support. A startup can pursue both — Bedrock POC first for the immediate feature work, then Accelerator admission for the larger AI roadmap — but the applications are filed through different paths.

VSOC 2 Type 2, CCPA, and the state-level privacy regimes — what US compliance looks like in 2026

The US compliance landscape for B2B SaaS startups is dominated by SOC 2 Type 2 — a voluntary attestation framework that is functionally mandatory for any startup selling to mid-market or enterprise B2B customers. Layered on top of SOC 2 are state-level privacy regimes (California, Virginia, Colorado, Connecticut, Utah, Texas, Oregon, Montana, Iowa, Delaware) that have proliferated since the 2018 California Consumer Privacy Act. Federal-level privacy legislation remains pending; the patchwork of state laws is the operative regime.

SOC 2 (Service Organization Control 2) is an attestation framework defined by the American Institute of Certified Public Accountants (AICPA), based on the Trust Services Criteria (Security, Availability, Processing Integrity, Confidentiality, Privacy). Type 1 attestation describes controls in place at a point in time; Type 2 attestation describes the operating effectiveness of controls over a defined observation period (typically 3–12 months). For B2B SaaS startups, Type 2 over a 6-month observation period is the typical first-attestation target; the audit is performed by a CPA firm specializing in SOC 2 (Prescient Security, Insight Assurance, A-LIGN, Schellman, BARR Advisory, Dansa D'Arata Soucia, and similar).

AWS publishes its SOC 2 Type 2 attestation through AWS Artifact (the customer console for compliance documents). The current AWS SOC 2 Type 2 report covers the AWS infrastructure layer — physical security, environmental controls, network security, change management, and the AWS-managed-service control surface. AWS customers consuming the infrastructure can reference the AWS SOC 2 in their own SOC 2 attestation as a "vendor management" control — the customer is responsible for SOC 2 controls in their application layer, but the AWS-layer controls are inherited from the AWS attestation.

The architectural pattern for SOC 2 Type 2 attestation on AWS, typical for a US B2B SaaS Series-A startup: AWS Organizations with separate accounts for production, staging, development, and security (typically 4–6 accounts at Series-A); AWS Config for compliance state tracking with conformance packs aligned to the Trust Services Criteria; AWS CloudTrail with multi-region trails and CloudTrail Lake for log analytics; AWS GuardDuty for threat detection; AWS Security Hub for security finding aggregation; AWS IAM Identity Center (formerly AWS SSO) for centralized access management with MFA enforcement; AWS Secrets Manager for credential management; AWS KMS for encryption-at-rest with customer-managed keys; VPC with private subnets and security groups; AWS WAF and Shield for application-layer protection; Amazon Inspector for vulnerability scanning. This architectural baseline is what Build for Startups credit applications scoped around "SOC 2 preparation" typically fund.

CCPA (California Consumer Privacy Act, in force since January 2020) and CPRA (California Privacy Rights Act, in force since January 2023, amending CCPA) establish consumer rights to know, delete, correct, opt-out of sale/sharing, and limit use of sensitive personal information for California residents. The architectural implications: consent management infrastructure (typically Amazon Cognito for user authentication with consent capture, AWS Lambda + Step Functions for consumer request workflows, Amazon S3 with Object Lock for retention compliance, AWS CloudTrail for access logging). For startups serving California consumers (which is most US-based B2C startups and a substantial share of US-based B2B SaaS), CCPA/CPRA compliance is operative regardless of where the startup is headquartered.

The state-level privacy regimes that followed CCPA/CPRA have created a patchwork of similar-but-not-identical statutes: Virginia Consumer Data Protection Act (VCDPA, in force since January 2023); Colorado Privacy Act (CPA, in force since July 2023); Connecticut Data Privacy Act (CTDPA, in force since July 2023); Utah Consumer Privacy Act (UCPA, in force since December 2023); Texas Data Privacy and Security Act (in force since July 2024); Oregon Consumer Privacy Act (in force since July 2024); Montana Consumer Data Privacy Act (in force since October 2024); Iowa Consumer Data Protection Act (in force since January 2025); Delaware Personal Data Privacy Act (in force since January 2025); plus newer 2025–2026 statutes in Tennessee, Indiana, New Hampshire, New Jersey, Minnesota, Maryland, Kentucky, Rhode Island. The architectural response is typically to implement the strictest combined requirements as a single posture — consumer rights workflows that comply with CCPA/CPRA also satisfy the other state statutes' lower-bar requirements.

For credit application narrative purposes, the standard US compliance language reads: "Production workload runs in us-east-1 across three Availability Zones. SOC 2 Type 2 audit in progress with [audit firm]; observation period ends [date]. AWS Artifact accessed for AWS SOC 2 Type 2 attestation referenced in vendor management controls. CCPA/CPRA compliance implemented through consent capture in Cognito, consumer request workflows in Lambda + Step Functions, retention controls via S3 Object Lock and CloudTrail access logging. State-level privacy regimes (VCDPA, CPA, CTDPA, UCPA, Texas DPSA, Oregon CPA) addressed through unified posture." This 4-sentence pattern is what CloudRoute partners pre-load into the US ACE record for B2B SaaS applications.

VIThe Activate Portfolio Sub-Program VC roster — and the partner-filed alternative when your fund isn't on it

The Activate Portfolio Sub-Program is the canonical path to $100K Activate credits for VC-backed US startups. The mechanism is straightforward: a VC enrolled in the Sub-Program has an LP allocation of Activate Portfolio credits, and the fund's platform team can submit a Portfolio request on behalf of any portfolio company. The credits land in the portfolio company's AWS account in 7–14 days. The Sub-Program-enrolled VCs are predominantly US-headquartered; the roster below is the working list as of 2026.

Andreessen Horowitz (a16z, Menlo Park / NY / SF): One of the longest-tenured Sub-Program VCs. a16z portfolio companies access $100K Portfolio through the a16z portfolio operations team. The a16z American Dynamism, Bio + Health, Crypto, Games, and Growth funds all participate. Typical wall-clock from a16z platform team submission to AWS credit balance: 7–10 days. a16z Speedrun (the early-stage accelerator) carries additional standing AWS benefits beyond the Portfolio path.

Sequoia Capital (Menlo Park, SF): Sequoia US portfolio companies access Portfolio through the Sequoia Arc team. Sequoia Arc (the accelerator) provides standing AWS benefits at admission. The Sub-Program enrollment is fund-level — Sequoia India and Sequoia Southeast Asia (now Peak XV Partners after the 2023 split) operate separate Sub-Program enrollments through different regional AWS teams.

Bessemer Venture Partners (Menlo Park, NY, Boston, Bangalore, Tel Aviv): Bessemer's cloud-focused funds (BVP X, BVP XII) have deep AWS engagement. Portfolio companies access $100K Portfolio through Bessemer's platform team. Bessemer's annual State of the Cloud report is widely-cited reference material in the broader AWS ecosystem.

Accel (Palo Alto, NY, London, Bangalore): Accel US portfolio companies access Portfolio through the Accel Atomic accelerator team and the broader Accel platform. Accel's cross-border investment activity sometimes routes Portfolio applications through different regional AWS teams depending on portfolio company HQ.

Greylock (Menlo Park, SF): Greylock portfolio companies — particularly the enterprise-software-focused funds — access Portfolio through the Greylock platform team. Greylock's portfolio support includes named AWS solutions architects for designated portfolio companies.

Founders Fund (SF): Founders Fund portfolio companies access Portfolio through the Founders Fund operations team. The fund's defense-tech and frontier-tech focus has produced increasing engagement with AWS GovCloud paths for relevant portfolio companies.

New Enterprise Associates (NEA, Menlo Park, DC, Chicago, Boston): NEA is one of the longest-tenured Sub-Program VCs with multi-stage participation. Portfolio companies across NEA's healthcare, enterprise tech, and consumer funds access $100K Portfolio.

Index Ventures (SF, NY, London, Geneva): Index US portfolio companies access Portfolio through the Index platform team. Index's cross-Atlantic structure means US-headquartered portfolio companies route through the US Sub-Program enrollment while London-headquartered companies route through EMEA.

Lightspeed Venture Partners (Menlo Park, NY, Bangalore, Tel Aviv): Lightspeed US portfolio companies access Portfolio through the Lightspeed platform team. Cross-border portfolio companies route based on HQ jurisdiction.

Insight Partners (NY): Insight's ScaleUp Society and the broader portfolio operations team handle Portfolio applications for Insight's growth-stage portfolio. Insight's late-stage focus means Portfolio is typically combined with MAP funding for larger workloads.

General Catalyst (Cambridge, SF, NY, Berlin / La Famiglia merger): General Catalyst US portfolio companies access Portfolio through the GC platform team. The 2024 La Famiglia merger extended GC's Portfolio access into Germany — German-headquartered LF portfolio companies now route through GC's US Sub-Program enrollment.

Tiger Global (NY), Coatue (NY, SF), Khosla Ventures (Menlo Park), GV (Mountain View): All Sub-Program-enrolled; portfolio companies access $100K Portfolio through respective platform teams. Tiger's and Coatue's growth-stage focus typically pairs Portfolio with MAP for larger migrations.

For US-incorporated startups backed by funds NOT on the Sub-Program-enrolled roster — including most micro-VCs, smaller seed funds, and family offices — the partner-filed ACE route is the alternative. Credit ceiling is identical ($100K Portfolio), wall-clock is slightly longer (11–18 days vs 7–14 for Sub-Program-direct), and the application requires a partner-attested ACE record rather than the VC platform team's Sub-Program submission. CloudRoute routes specifically to APN Partners with documented Sub-Program-parity track records in the US Americas queue.

VIIYC, Techstars, 500 Global, Antler, On Deck, AI Grant, a16z Speedrun, Sequoia Arc — the US accelerator dense layer

The density of US-based accelerators with standing AWS Activate offers is one of the structural advantages US startups have over international peers. A US founder with admission to any of the major accelerators carries at least one standing AWS credit offer plus a credibility signal that lifts subsequent partner-filed applications. The roster below covers the accelerators that consistently appear in CloudRoute's routed US pipeline.

Y Combinator (Mountain View; W and S batches with ~250 companies per batch): YC carries $5K standing AWS Activate credits via the YC startup-services agreement. The mechanism: admission to a YC batch triggers an email with the AWS Activate enrollment link pre-attributed to YC. The credits apply within 24–72 hours. Beyond the standing offer, YC graduates have direct access to the Portfolio Sub-Program path — most major Sub-Program VCs treat YC graduation as a sufficient institutional signal even for companies that have not yet raised institutional capital from the Sub-Program VC itself. YC is also notable for its AI-focused offerings (YC AI Startup School, YC AI Demo Day) which connect graduates to the Bedrock POC and Generative AI Accelerator pipelines.

Techstars (US-based accelerators in Boulder, NY, Boston, Chicago, Austin, LA, Seattle, Atlanta, DC, Detroit, and sector-specific verticals): Techstars carries $1K standing AWS Activate credits with documented escalation paths through the Techstars partner perks portal. Beyond the standing credits, Techstars participation is recognized as an accelerator signal in partner-filed Founders applications, typically lifting approvals toward the $20K–$25K end of the range.

500 Global (formerly 500 Startups, SF): 500 Global carries $5K standing AWS Activate credits and accelerator-program signals. The 500 Global Founders Series, Flagship Accelerator, and sector-specific programs (500 Spartans for Saudi, 500 RWA for tokenization) all carry the standing offer.

Antler (selective in US, with Antler New York, Antler West, Antler Austin): Antler's US programs are selective — typical Antler US cohort sizes are 20–30 founders versus 250+ for YC. Antler portfolio companies typically access $5K standing AWS Activate plus accelerator-signal lift on partner-filed applications. Antler's cross-border structure (Antler Berlin, Antler London, Antler Singapore, etc.) means US-incorporated Antler companies route through US-side mechanics.

On Deck (SF, NY, fellowship-based): On Deck operates fellowship-based programs (ODF1, ODF2, etc. plus sector-specific cohorts). On Deck participation is recognized as an accelerator-adjacent signal in partner-filed applications. The fellowship structure doesn't carry a standing AWS offer in the same way batch accelerators do, but the On Deck network has produced numerous Sub-Program VC-backed companies.

AI Grant (SF, AI-native focused): AI Grant is a more recent (2022-founded) AI-focused fellowship that has produced notable AI-native portfolio companies. AI Grant participation is increasingly recognized in Bedrock POC application narratives — the AI-native specialization aligns with Bedrock's product focus.

a16z Speedrun (SF, gaming and consumer focus): Speedrun is a16z's early-stage accelerator with a gaming/consumer-tech focus. Speedrun participation carries Sub-Program access for a16z portfolio mechanics. Typical Speedrun cohort sizes are 20–40 companies per cohort.

Sequoia Arc (SF, NY, London): Arc is Sequoia's early-stage accelerator program with cohort-based investment. Arc participants access Sequoia's Sub-Program Portfolio path directly. Cohort sizes are 15–25 companies per cohort.

Pear VC Accelerator (Palo Alto), Plug and Play Tech Center (Sunnyvale, multi-vertical), Mucker Capital (LA), Founders Inc (SF), Entrepreneur First (NY, London, Bangalore, Singapore, Berlin, Paris), South Park Commons (SF, NY): All produce accelerator-signal recognition in partner-filed applications. South Park Commons specifically is a community-based program (no batch structure) but the SPC network has produced enough Sub-Program-VC-backed companies to be recognized as a credibility signal.

The structural effect of accelerator density: most US founders have at least one accelerator signal available — even if not formally activated. CloudRoute observation: roughly 65% of US-incorporated startups in the routed pipeline have either current or past accelerator affiliation, against 30–40% for international startups. The accelerator-signal lift on partner-filed Founders applications is typically $5K–$10K higher approval, which is the difference between landing at $15K and landing at $25K.

VIIIBedrock POC, the Generative AI Accelerator, and the US share of accepted cohorts

The AWS Generative AI Accelerator is the largest AI-specific credit program AWS operates. Inaugurated in 2023 with the first cohort, the program has run multiple cohorts through 2025 with US-headquartered startups representing the plurality of accepted companies. The $300K median credit allocation (with a $1M ceiling for the largest-funded participants) is materially above the Bedrock POC pool ($10K–$50K) — but the admission process is competitive and the application requirements are substantial.

The Generative AI Accelerator is operated by the AWS Startup team in conjunction with the Amazon Bedrock product team. Cohort admission is by competitive application — typical acceptance rates have been in the single-digit percent range for the inaugural 2023 cohort, with somewhat higher rates (low double-digit percent) for the 2024–2025 cohorts as the program scaled. The 2024 cohorts expanded geographic eligibility to include EMEA and APAC applicants more substantively; US applicants remained the plurality of accepted companies through 2025 due to AI-native company density in the SF and NY ecosystems.

The Accelerator engagement window is typically 10 weeks of structured programming (one week orientation, eight weeks of working sessions with AWS solutions architects, one week of demo presentations to AWS and partner investors). The credit allocation is a designated credit pool of $300K median (with documented upside to $1M for the largest-funded participants and accepted-cohort-stage Series-B+ companies). Beyond the credits, participants receive dedicated AWS solutions architect engagement, Bedrock product team office hours, foundation-model-vendor access (Anthropic, Meta, Mistral, Amazon Bedrock model teams), and an investor-network introduction layer.

The application requires: a defined use case for foundation-model-based AI within the startup's product; documented evaluation methodology for the use case (typical evaluation involves a held-out test set with quantitative metrics — accuracy, latency, cost per inference); a 10-week roadmap aligning the Accelerator engagement with the product roadmap; team commitment from at least one engineering lead and one product lead; existing AWS account in good standing.

For Bedrock POC funding (the separate, smaller pool of $10K–$50K), the application is partner-filed via ACE rather than direct competitive application. The pool is more accessible — most partner-filed Bedrock POC applications with a defined scope and evaluation methodology approve at $25K–$50K. The Bedrock POC pool is Bedrock-earmarked (the credits apply specifically to Bedrock inference and Bedrock-adjacent service consumption: Bedrock model inference, Bedrock Knowledge Bases, Bedrock Agents, Bedrock Guardrails, plus the vector database services like Amazon OpenSearch Service and Amazon Aurora pgvector that typically pair with Bedrock RAG patterns).

For US-incorporated AI-native startups, the strategic play is typically to pursue Bedrock POC first (faster wall-clock, partner-filed, smaller pool but accessible) and apply for the Generative AI Accelerator separately for the larger pool and the structured engagement. The applications do not interfere with each other — a startup can hold both an active Bedrock POC credit balance and a pending Accelerator application. CloudRoute observation: roughly 35% of US Series-A AI-native applications in the 2025–2026 routed pipeline pursued both paths in parallel.

Model availability in us-east-1 and us-west-2 is the broadest of any AWS region globally. Frontier model releases (new Claude Sonnet versions, Llama 3.x versions, Mistral Large updates, Amazon Nova updates) typically launch in us-east-1 first, often on the same day as the model vendor's public announcement. us-west-2 follows in 0–14 days; other regions follow over the subsequent 30–90 days. For AI-native US startups requiring latest-model access, us-east-1 or us-west-2 is the operational requirement.

IXThe US Partner-Filed application narrative — exactly what the partner writes into ACE

Every Partner-Filed Activate application is an ACE record. The record has structured fields (company info, use case, AWS services, projected spend) and a free-text narrative section. The narrative section for US applications carries the SOC 2 / CCPA / state-privacy compliance language and the institutional-signal references (VC backing, accelerator affiliation, Delaware C-corp confirmation). Here is the structural pattern a CloudRoute-routed US partner uses for a typical Series-A application.

Company-info block (4 sentences): "New York-headquartered Series-A B2B fintech SaaS, 18 engineers, $14M Series-A closed Q4 2025 led by Bessemer Venture Partners with Index Ventures and Accel participation. Delaware C-corp (formation July 2023, Delaware File Number [number]); registered in New York for business operations; primary banking with Mercury Bank. Building a payment-reconciliation and treasury-management SaaS for mid-market B2B finance teams. Existing AWS spend $4.2K/month on a partially-built us-east-1 footprint."

Use-case paragraph (Portfolio): "Production workload runs in us-east-1 (Northern Virginia) across three Availability Zones for HA, with us-east-2 (Ohio) configured as a multi-region DR target for resilience. Service mix: EKS for the application control plane, Amazon RDS Aurora PostgreSQL for transactional data, Amazon DynamoDB for high-volume ledger entries, Amazon S3 with Object Lock for retained transaction records (7-year retention per financial-services norms), Amazon Kinesis Data Streams for transaction event streaming, AWS Lambda for the reconciliation processing pipeline, Amazon Cognito for B2B SSO via SAML/OIDC, AWS Step Functions for the workflow orchestration. AWS Artifact accessed for AWS SOC 2 Type 2 attestation. Projected AWS consumption: $8K–$12K/month at end of 2026 ramp."

Use-case paragraph (Build for Startups, distinct workload): "Distinct from the production workload above: we are deploying SOC 2 Type 2 preparation infrastructure across the AWS Organization in advance of our Q3 2026 audit observation period commencement. Scope: AWS Config with SOC 2 conformance packs across all 5 accounts in the Organization; AWS Security Hub aggregating findings from GuardDuty, Inspector, and Macie; AWS CloudTrail Lake for centralized log analytics with 1-year retention; AWS IAM Identity Center for centralized access management with MFA enforcement; AWS Systems Manager for patch management; vulnerability scanning via Amazon Inspector with automated remediation runbooks. Projected consumption for this discrete project: $1.8K/month. Audit firm engagement [audit firm name]; observation period commences [date]."

Use-case paragraph (Bedrock POC, AI workload): "Adding a customer-support copilot to the payment-reconciliation product: an LLM-assisted reconciliation-anomaly-investigation tool that ingests transaction history, retrieval-augments against the customer's payment-rules knowledge base, and produces structured anomaly-classification hypotheses for the finance team to review. Model selection: Claude Sonnet 4 (us-east-1) for the primary reasoning, Amazon Titan Embeddings for the RAG layer over the payment-rules KB. Evaluation methodology: N=300 historical reconciliation anomalies with verified classifications; accuracy measured as top-1 classification match against ground-truth; latency target sub-3-second response; weekly cadence over the 60-day POC window. Projected Bedrock spend: $2.2K/month at POC scale."

Compliance addendum (US market-specific): "SOC 2 Type 2 audit observation period commences Q3 2026; AWS Artifact accessed for AWS SOC 2 Type 2 attestation referenced in vendor management controls. CCPA/CPRA compliance implemented through Cognito consent capture and Lambda + Step Functions consumer request workflows; State-level privacy regimes (VCDPA, CPA, CTDPA, UCPA, Texas DPSA, Oregon CPA) addressed through unified posture. Financial services-adjacent posture maintained — workload designed for future SOC 1 Type 2 and PCI-DSS scope expansion when scope warrants. AWS Organizations SCP enforces region restriction to us-east-1 and us-east-2 only for production accounts; AWS GuardDuty enabled across all accounts; CloudTrail centralized; AWS Backup configured with 90-day retention."

XThe end-to-end engagement workflow for a US Series-A startup

A typical end-to-end engagement for a US-incorporated Series-A startup, from initial inquiry to credits applied to the AWS account. Wall-clock timing pulled from CloudRoute's routed US pipeline through 2025–2026.

Day 0: Inquiry submitted to CloudRoute (3 minutes). Routing to a US Americas-queue-experienced APN Partner happens within 24 hours. Partner selection is based on vertical (B2B SaaS, fintech, dev-tools, B2C, AI-native), city (SF, NY, Austin, Boston, Seattle, LA, Chicago, Denver, Miami), VC backing (Sub-Program VC vs non-Sub-Program), and compliance scope (SOC 2 prep, HIPAA, PCI-DSS, FedRAMP).

Day 1–2: 30-minute discovery call. Partner confirms Delaware C-corp status (or other incorporation), AWS account in good standing, VC backing and Sub-Program enrollment status, accelerator affiliations, current AWS spend, projected workload scope, compliance roadmap, AI roadmap. The discovery is structured around the three credit-pool framing: Portfolio (production workload), Build for Startups (distinct second workload, typically SOC 2 prep or second product line), Bedrock POC (AI workload with defined evaluation methodology).

Day 3–5: Founder provides company info (4 sentences company-info block), AWS account ID, use case descriptions (1 paragraph per pool), VC partnership letter or accelerator confirmation (1 page typical), 8–10 slide deck. For Bedrock POC scope, the founder also provides the evaluation methodology (held-out test set, accuracy metric, latency target, cost target).

Day 5–7: Partner files ACE records — typically three records: Portfolio ($100K), Build for Startups (+$25K), Bedrock POC (+$25K). Each record carries the US compliance addendum, Delaware C-corp confirmation, institutional-signal references (Sub-Program VC or accelerator affiliation), and the projected consumption itemized by AWS service.

Day 7–14: US Americas queue assigns. Sub-Program-parallel applications (with explicit Sub-Program VC backing references) typically clear in the lower end of this window. Non-Sub-Program applications clear in the middle of the window. The reviewer may ask one clarifying question about region selection, compliance scope, or projected consumption; partner responds within 24 hours.

Day 11–18: Credits applied to the AWS account, visible in the Billing and Cost Management dashboard under "Promotional credits." Portfolio credits ($100K) are typically a single credit code valid for 24 months; Build for Startups credits ($25K) are typically a separate code valid for 12 months; Bedrock POC credits ($25K) are typically a Bedrock-earmarked code valid for 12 months with consumption restricted to Bedrock and Bedrock-adjacent services.

Total founder time: ~2–3 hours across the engagement. Total wall-clock: 11–18 days from inquiry to credits applied. Total cost: $0 — AWS funds the partner via APN Funding and ACE attribution; the partner pays CloudRoute commission from its own AWS-funded revenue.

XISix mistakes US founders make on AWS credit applications

Mistake 1: Claiming self-serve before checking the auto-issuance handoffs. Stripe Atlas, Mercury, Brex, and Ramp all carry $5K auto-issuance Activate Founders credits. The credits don't stack — only the first claim applies. A founder who self-serve-claims $5K Activate before checking the Atlas dashboard has potentially missed nothing (the credit is the same $5K) but has also missed the lower-friction path. The fix: check the auto-issuance handoffs before self-serve.

Mistake 2: Waiting for the VC to file Portfolio when partner-filed is faster. Sub-Program VCs file Portfolio applications, but the timing varies — some Sub-Program VCs file within days of the Series-A close; others batch applications quarterly. The published timeline is 7–14 days but the wall-clock at the founder's end can be 30–90 days depending on the VC's submission cadence. The partner-filed ACE alternative reaches the same $100K ceiling in 11–18 days reliably. The fix: ask the VC platform team for a wall-clock estimate; if it's longer than 2 weeks, engage a CloudRoute partner in parallel.

Mistake 3: Filing Build for Startups for the same workload as Portfolio. Build for Startups requires a distinct workload — a second product line, a separate compliance build-out, a discrete migration project, or a separate experimental product. Filing Build for the same production workload as Portfolio leads to reviewer clarification (and typically denial of the Build pool, retaining only the Portfolio $100K). The fix: scope Build for Startups around the SOC 2 prep, the second product, or the separate migration explicitly with named AWS services that don't overlap with the Portfolio scope.

Mistake 4: Scoping Bedrock POC without an evaluation methodology. The Bedrock POC pool requires a defined evaluation methodology — a held-out test set, a measurable accuracy or quality metric, a latency or cost target, a 60–90 day evaluation window. Applications without this scoping typically land at the floor ($10K) or get clarification cycles. The fix: provide a 4-sentence evaluation methodology block with the application — held-out N, accuracy metric, latency target, cost target.

Mistake 5: Defaulting to us-east-1 without checking for service-availability dependencies. us-east-1 has the broadest service catalog, but some specialized services have historically launched in us-west-2 first or have us-west-2-only availability. AWS Local Zones in some metros are tied to specific parent regions. For specialized workloads (specific ML accelerator instance families, specific managed services), confirming the parent-region availability for the workload before defaulting to us-east-1 avoids re-architecture friction. The fix: cross-reference the workload's required service list against the AWS Regional Services List before finalizing region selection.

Mistake 6: Treating the Generative AI Accelerator as a Bedrock POC. They are different programs with different application paths. Bedrock POC is partner-filed via ACE for $10K–$50K with 11–18 day wall-clock. The Generative AI Accelerator is a competitive cohort application with cohort-tied admission timing ($300K median, $1M ceiling) and 10-week structured programming. Submitting a Bedrock POC scoped application to the Generative AI Accelerator (or vice versa) creates routing friction. The fix: pursue Bedrock POC for immediate feature work and the Accelerator separately for the larger AI roadmap.