- Microsoft is investing $2.5 billion and embedding 6,000 employees into Microsoft Frontier Company to implement AI inside client businesses.
- Amazon, OpenAI and Anthropic have launched comparable ventures since May 2026, worth a combined $6.5 billion.
- The shift puts funded startups such as Riplo, Xavier AI, Duvo, and Lyzr AI in competition with, or dependence on, the same companies whose models they build on.
Microsoft is putting $2.5 billion and 6,000 employees behind a new business called Microsoft Frontier Company, built to send engineers directly into client companies to get artificial intelligence actually working. It is not selling more software. It is selling the thing a growing wave of venture-backed startups was funded to sell: implementation.
The new unit, led by former Microsoft Asia president Rodrigo Kede Lima, folds together forward deployed engineers, consultants and technical specialists into a single organisation.
Judson Althoff, chief executive of Microsoft’s commercial business, wrote that the effort “goes beyond what has been labeled as Forward-Deployed Engineering” and is designed to be the industry’s largest, most outcome-driven engineering organisation. In Microsoft’s own announcement, Althoff frames the unit around two ideas he calls “Intelligence + Trust”: helping a company’s proprietary data and workflows compound into a genuine advantage, while guaranteeing that data is never used to train models in ways that erode what makes the company distinctive in its market.
Microsoft named four early customers in its announcement: the London Stock Exchange Group, where Frontier engineers embedded AI into LSEG Workspace so finance professionals can query structured and unstructured financial content directly, along with Land O’Lakes, Unilever, and Novo Nordisk.
Microsoft says the platform is deliberately model-agnostic, letting customers run OpenAI, Anthropic, Microsoft’s own models or open-source alternatives rather than locking them into one vendor.
Four Big Tech companies are in
Microsoft is not alone, and it is not even first. Amazon Web Services committed $1 billion to its own forward deployed engineering push two days before Microsoft’s announcement.
OpenAI and Anthropic moved earlier still: OpenAI stood up a standalone Deployment Company in May 2026 backed by more than $4 billion from a TPG-led investor group, while Anthropic partnered with Goldman Sachs, Blackstone, and Hellman & Friedman on a $1.5 billion venture that embeds engineers inside mid-sized companies, starting with the investment firms’ own portfolios.
Four of the best-capitalised companies in AI have independently reached the same conclusion within a matter of weeks: the money is no longer just in building models, it is in making them work inside a real company’s messy, specific systems.
The forward deployed engineering model itself is not new. Palantir pioneered it roughly two decades ago, sending its own engineers to work alongside the U.S. military, but 2026 is the year every major AI lab and cloud vendor decided to run the playbook simultaneously, at a scale none of them attempted before.
The startups sitting in the collision path
That conclusion should worry a specific category of startup: the ones raising seed and Series A rounds to do exactly this.
London-based Riplo raised £2.3 million from Cherry Ventures to build what it calls an operating system for consulting firms, aiming to replace slide decks with a structured AI workspace for managing client deliverables. Xavier AI launched out of stealth aiming to raise $15 million for an AI strategy consultant, drawing an explicit line between strategy work and implementation work, the exact seam Microsoft Frontier Company now proposes to occupy with 6,000 people and a global systems-integrator network attached, including launch partners Accenture, Capgemini, EY, KPMG, and PwC.
Not every startup in the category sits on the losing side of this. Duvo, the Prague and New York-based retail automation startup founded by Rohlik’s Tomas Čupr, raised $15 million partly by positioning itself against large consulting projects that stitch together legacy systems. And Lyzr AI, which hit a $250 million valuation with Accenture leading its round, has built its business selling infrastructure to the consultancies themselves, including Accenture, Deloitte, and KPMG.
The bigger question for founders
The pattern underneath all of it is familiar from software’s last major platform shift, compressed into a far shorter timeline: a genuinely hard, unglamorous problem becomes valuable enough that the biggest incumbents decide to own it directly rather than buy it from a startup or a systems integrator.
For founders building in this space, the question is no longer whether Big Tech will compete with them. It already is, at a scale most Series A companies cannot match, and with a distribution advantage that a startup has to work years to build. The harder question is whether the pitch to investors still holds once the customer can get a comparable service bundled in with the infrastructure they were already paying for.
