From $134M Loss to $6.3B Acquisition: How AI Made Amex GBT Profitable

For twelve years, Amex GBT was the largest standalone corporate travel platform on earth and a money-losing public company. In 2024 it posted a $134 million net loss. Yesterday it agreed to be taken private for $6.3 billion at a 60% premium, with applied AI on the cover of the announcement.

The arc between those two facts is the most important data point in B2B travel in a decade.

The Profitability Inversion

Eight weeks before the deal, Amex GBT reported its first profitable year as a public company: $111 million in net income for 2025. Management led the earnings call with AI, not with the CWT acquisition that explained most of the headline number.

Over five years, digital transactions moved from 60% to 83% of volume, and Egencia's average booking time dropped under three minutes. The next-generation Egencia, launching this April, replaces forms with natural-language requests routed through AI agents that follow policy automatically. This is what a margin re-architecture looks like — not a feature, but a slow rebuild of the cost base.

Why This Was Always the Hard Problem

Corporate travel has been brutal to run profitably at scale. Service obligations are heavy and human-intensive, content is fragmented across GDSs, NDC, and direct connections, and travel buyers have proven structurally resistant to higher TMC fees.

Every wave of B2B travel technology in the last twenty years promised to break that vice and produced incremental savings instead. AI is the first technology that has actually compressed the human-cost layer without breaking the service promise — and the proof points are now industry-wide. Flight Centre's corporate brands recently reported a 20% increase in transaction value per employee over two years, attributed explicitly to AI tooling.

The Capital Market's Verdict

Long Lake Management was founded in 2023. Its prior portfolio consisted of eighteen homeowners-association management companies. Yesterday it agreed to acquire a 22,000-person, $2.7 billion-revenue global travel platform for $6.3 billion in cash.

The equity came from General Catalyst, Alpha Wave, and Koch — the same investors backing OpenAI and Anthropic. The capital that built the AI labs is now financing the travel software those labs will run on.

The capital that built the AI labs is now financing the travel software those labs will run on.

The thesis would have been laughed at three years ago: corporate travel — bookings, changes, cancellations, expense reconciliation — is a high-volume, partially deterministic workflow with strong policy guardrails. That description matches what current-generation AI agents now handle competently.

What It Means for Everyone Else

The post-deal landscape is no longer organized around whether AI matters in business travel. It is organized around who runs the AI rebuild at scale, who gets absorbed into one, and who builds something structurally different.

For mid-sized TMCs, competing with a $6.3 billion AI-rebuild thesis on its own terms is not a fight worth picking. Specializing in segments where human expertise is the product is the obvious defensible path — but it leaves the volume base of corporate travel exposed. For travel agencies and smaller operators, the choice is no longer whether to build on AI, but whether to build on someone else's infrastructure or remain a manual operation in a market that has moved on.

For investors, the signal is unambiguous: the largest discretionary capital pool in business travel just declared a winner-take-most market. Profitable business travel through AI is no longer a speculative thesis. It is now the operating premise.

For ongoing coverage of the AI rebuild of travel infrastructure, subscribe to Bitravel.