On-Premises AI Infrastructure company raised $85 million as banks accelerate deployments of private AI infrastructure within corporate data centers.

Go.AI, the Chicago company that sells private artificial intelligence systems to banks and other regulated institutions, has raised an $85 million Series A, signalling that on-premises AI infrastructure is maturing into a paying market rather than a compliance workaround. The round, announced on 22 September 2026 and led by Washington DC growth-equity firm Updata Partners, takes the company's total funding to $90 million. Existing backers GFT Ventures and LAUNCH also participated.

For finance leaders, the relevant point is not the headline number but what buyers are paying for: an AI stack—software and hardware together—that runs entirely inside a customer's own secure environment, so sensitive records never leave the building. The company reached this milestone three weeks after retiring its former name, Go Abacus, and moving to the go.ai domain on 1 September.

The Series A was led by Updata Partners, a technology-focused growth-equity firm with more than $3 billion in committed capital, with general partner Carter Griffin taking the deal. GFT Ventures and LAUNCH, both existing shareholders, followed on. This financing builds on a $5 million seed round in November 2025 that GFT Ventures led, with BankTech Ventures and LAUNCH also participating. The company reported roughly 50 employees at the time of the raise and plans to use proceeds to expand its engineering team, accelerate development of its Go.OS software operating system and hardware line, and fund expansion beyond regulated sectors toward a wider set of compliance-minded organisations.

Banks have spent much of the past decade wary of putting core workloads in the public cloud, citing data residency, supervision and third-party risk. That caution is now repeating with public large language models, where the sticking point is that using a hosted model can mean sending customer, patient or member data to an external provider. For an institution operating under bank secrecy, prudential and audit obligations, that is a licensing and control problem before it is a technology choice. Go.AI's design answers that objection directly: its systems run models and index information within the customer's own environment, with no proprietary data passed to third parties.

The founders' background reinforces this focus. Lisa Gillespie, co-founder and chief operating officer, spent more than two decades teaching and practising accounting and taught chief executive David Moscatelli at Loyola University Chicago before the pair started the business as Abacus Analytics LLP in 2018. The company became Go Abacus Corporation in 2022 and positions its infrastructure as auditable and ready for regulatory examination, a framing aimed squarely at compliance and internal-controls teams.

The commercial proposition most likely to resonate with chief financial officers is pricing. Go.AI charges a fixed fee and does not bill per token, unlike the usage-based model common to hosted AI services. As query volumes rise, per-token pricing turns AI into a variable cost that is hard to forecast—a live concern for any institution moving from pilots to production. Go.AI's own customers, on the company's figures, process more than 12.5 million queries a day, a volume at which usage-based billing compounds quickly. The proposition is therefore closer to an infrastructure purchase than a software subscription: fixed, capital-style economics in exchange for owning the deployment. That trade appeals to buyers who need budget certainty and control.

Several of the company's performance figures are self-reported and should be read as such. Go.AI states that it serves more than 200 customers across financial services, healthcare, aerospace and defence, and manufacturing; that annual recurring revenue has risen more than eightfold year on year; and that the business is profitable. These metrics, disclosed by the company and reported by Axios, have not been independently audited, and Go.AI has not published absolute revenue figures. The profitability claim is particularly noteworthy: positive cash generation is uncommon among AI infrastructure companies raising at this stage, most of which are burning capital to buy growth, so if it holds under scrutiny it is a genuine differentiator. One product detail also merits scrutiny by prospective buyers: Go.AI's Go.OS documentation describes local inference by default with optional cloud routing for selected tasks, while its banking materials describe fully on-premises deployments with no cloud fallback or external calls. Regulated buyers will want the deployment they are sold specified in the contract.

The company's flagship appliance, The Go1, launched on 24 March 2026 under the Go Abacus name and packages computing hardware with Go.OS. Go.AI describes it as the first on-premises AI hardware and software product built for regulated organisations; that claim reflects the company's own positioning and has not been independently verified.

Go.AI is entering a field filling from two directions. The large cloud providers offer private and virtual-private-cloud deployments intended to reassure regulated customers, while a separate set of vendors sells sovereign or self-hosted AI as software that runs on the customer's own kit. Go.AI's distinction is that it supplies the hardware as well as the software as a single appliance, priced at a fixed fee. Whether that integrated, capital-style model out-competes software-only sovereign platforms and the incumbents' private-cloud options is the commercial question the funding is meant to answer.

The signal for finance professionals is that private, on-premises AI has moved from a compliance concession to a fundable business with real revenue behind it. Go.AI's raise, its stated profitability and its regulated-industry customer base suggest demand for AI that never leaves the institution is substantial, not marginal. For chief financial officers, the fixed-fee model reframes AI as a controllable line item rather than an open-ended usage bill, a distinction that becomes sharper as query volumes scale. For investors, a cash-generative infrastructure company in a category dominated by cash-burning peers is a rare profile, provided the self-reported numbers survive due diligence. The strategic contest over the coming capex cycle is whether owning the stack outright beats renting compliance from the cloud.