Go.AI Raises $85M for On-Prem AI Infrastructure

The buyer for enterprise AI changes when an examiner can ask where every answer came from. A fast demo still matters, but so do the data path, the audit trail, the deployment boundary and the person who will have to defend the system after the vendor leaves the room.

Go.AI has raised an $85M Series A led by Updata Partners to build for that buyer. Existing investors GFT Ventures and LAUNCH also participated. The Chicago company said the round brings total funding to $90M and will expand engineering, accelerate its Go.OS software and Go1 hardware roadmap, and scale go-to-market efforts beyond regulated industries into a wider set of compliance-minded organizations.

What Go.AI Announced

Go.AI announced the financing on September 22, 2026. Updata Partners VII led the round, and Updata General Partner Carter Griffin will join the board. The company did not disclose its valuation, ownership terms, investor allocations or the size of each participant's check.

The Series A follows a $5M Seed round announced in November 2025. GFT Ventures led that financing, with BankTech Ventures and LAUNCH Fund participating. The two disclosed rounds reconcile with Go.AI's stated $90M total, although the company has not published a full capitalization history or detailed financing terms.

The corporate timeline deserves care. Go.AI's current history says David Moscatelli and Lisa Gillespie founded Abacus Analytics LLP in 2018, converted and rebranded it as Go Abacus Corporation in 2022, then adopted the Go.AI name on September 1, 2026. An older seed announcement described the business as founded in 2022. The public record supports 2018 as the start of the predecessor partnership and 2022 as the corporate conversion and product refocus, rather than pretending the two dates mean the same thing.

The Product Is an Operating Boundary

Go.AI sells private AI infrastructure to banks, credit unions, healthcare organizations, insurers, aerospace and defense companies, manufacturers and other organizations with strict security or compliance requirements. Its stack combines the Go1 appliance with Go.OS, which manages models, data indexing, applications and audit records inside the customer's environment.

The company says a standard Go1 can serve as many as 2,000 concurrent users, move from network connection to production in about 15 minutes, and run without a public-cloud dependency. It offers fixed pricing rather than per-token billing. Those specifications come from Go.AI's product materials and have not been independently benchmarked in the public record reviewed for this article.

Updata describes the system as deployable on customer-managed hardware or within a private-cloud instance. Go.AI's own product pages emphasize on-premises and air-gapped configurations, while Go.OS documentation says customers may choose to route selected tasks to cloud models under the same audit chain. The useful distinction is control: the architecture is designed to let the institution decide which workloads, models and data paths remain local rather than accepting one cloud policy for every task.

Why Regulated AI Is a Different Sale

In a conventional software sale, product capability and return on investment can carry most of the conversation. Regulated AI introduces additional buyers and veto holders. Security teams care about data egress, compliance leaders need records, model-risk teams need governance, IT has to operate the system, and examiners may later ask how a particular output was produced.

That makes the deployment boundary part of the product. Go.AI is betting that organizations will pay for owned capacity, local control and predictable costs even while the broader AI market keeps moving toward shared cloud infrastructure and variable consumption. The pitch becomes especially relevant when sensitive documents, customer records or clinical information cannot casually travel through third-party systems.

The tradeoff is that hardware and local infrastructure create their own obligations. Appliances must be installed, secured, updated, monitored and supported across customer environments. Go.AI's current careers page reflects that operational reality with openings spanning client advisory, field deployment, sales engineering, marketing and strategy rather than only model research.

The Growth Claims and Their Limits

Go.AI reports more than 200 customers, annual recurring revenue growth above 8x year over year, continued profitability and more than 12.5M queries processed daily across customer deployments. Its website also says the platform serves 900,000 monthly users. Updata repeated the ARR, profitability and query figures in its own investment announcement.

These are material signals, but they remain company and investor claims. Go.AI has not disclosed the prior-year ARR base, current revenue, contract values, customer concentration, gross margin, retention, hardware deployment count or audited profitability. A customer count also does not show how broadly each account has deployed the product, and a query total does not reveal the business value or risk profile of those queries.

The company says it has grown from its 2 founders to more than 50 team members. Its LinkedIn page lists a 51-to-200 employee range and a headquarters at 111 South Wacker Drive in Chicago, where the company opened a new office in July 2026 with initial capacity for 50 employees. Those facts support a visible operating expansion, while the undisclosed financial details keep the quality and durability of that growth outside public verification.

What the $85M Has to Prove

The capital gives Go.AI room to expand engineering, develop the Go.OS and Go1 product families, hire client-facing teams and sell beyond its original financial-services base. It also changes the standard applied to the company. A large Series A can fund faster product development, but customers in regulated markets judge infrastructure through reliability, controls, support and the quality of the evidence available during an audit.

Go.AI now has to show that its local-first architecture can scale across more institutions without turning every deployment into a custom integration project. It must keep hardware supply, software updates, security controls, model governance and client advisory moving together while preserving the simplicity promised by a box that can be installed quickly.

That is the larger signal inside the round. The next enterprise AI market will not be divided only by which model answers best. It will also be divided by who can explain where the model ran, what it touched, what it cost and who remains accountable when the answer enters a regulated workflow. Go.AI has raised $85M to make those questions part of the infrastructure instead of an argument that begins after deployment.