Summary

I've been very surprised that the market treats OpenAI's latest partnership hire as proof that the company is pivoting from growth-at-all-costs to disciplined commercialization. It isn't. What it is, is a structural admission: OpenAI has $25 billion in annualized revenue but no enterprise sales engine of its own, so it's borrowing one from Salesforce by the person who built it.

Brian Landsman spent 14 years at Salesforce, most recently as executive vice president of global partnerships and CEO of AgentExchange - Salesforce's partner ecosystem and app marketplace. He leaves a company that sells enterprise software profitably to join a private company that projects $14 billion in losses for 2026. That is not a maturation play. That is an urgency play.

The false narrative here is that Landsman's arrival signals OpenAI is finally solving its go-to-market problem and inching toward sustainable cash generation. In my opinion, it signals something different: OpenAI knows that revenue growth alone cannot justify its $852 billion valuation, and it needs to build an enterprise channel fast because its burn rate is the fastest of any startup in history.

Let's look at the numbers that matter - the ones this narrative doesn't change.

OpenAI generated $2 billion in revenue per month as of early 2026, up from $1 billion per quarter by the end of 2024. Enterprise AI is already roughly 40% of its revenue mix. By the end of 2025, annualized revenue had climbed from $20 billion to $25 billion in February. The growth is real.

However, OpenAI's burn rate is projected at 57% of revenue in 2026 and 2027, and internal documents show $14 billion in losses for 2026 alone - nearly triple its 2025 losses. Over the 2023 through 2028 period, OpenAI expects to lose approximately $44 billion before turning a profit of $14 billion in 2029. That is a six-year hole that the company filled with a $122 billion funding round closed in March 2026 at an $852 billion post-money valuation.

That being the case, the structural question isn't whether Landsman can build partnerships. It's whether any partnership structure can bridge the gap between a $14 billion annual burn and a path to free cash flow positive operations within a reasonable timeframe.

Here is where the AI cross-pollination model becomes relevant. OpenAI's spending doesn't disappear. It flows downstream to the infrastructure layer, and that is where the free cash flow actually sits.

Consider Broadcom. Earlier this year, OpenAI's custom AI chip deal with Broadcom expanded to a scope that hit an $18 billion financing threshold. Broadcom's AI revenue surged 106% year-over-year to $8.4 billion in its first quarter of fiscal 2026. Annual free cash flow was $26.9 billion in 2025. The company pays a $0.65 quarterly dividend and returns capital systematically. Broadcom builds the hardware that makes OpenAI's models run, and Broadcom gets paid in cash with every order.

Now consider Nvidia. Fiscal 2026 free cash flow was approximately $119 billion for the year ending in April. Nvidia returned $41.1 billion to shareholders in fiscal 2026 through buybacks and dividends, and CEO Jensen Huang committed to directing 50% of free cash flow in 2026 toward share repurchases and dividends. Nvidia supplies the GPUs that power the compute clusters OpenAI is building.

Then there's Microsoft, which has generated more than $30 billion in revenue from its Azure OpenAI services alone. Microsoft just ended its three-year exclusive cloud partnership with OpenAI in April 2026, capping the revenue share it pays back and giving OpenAI freedom to work with AWS and Google Cloud. The structural result: Microsoft keeps its existing revenue stream, sheds the payout obligation, and retains first-ship rights on Azure. Microsoft wins either way.

The pattern is clear. OpenAI spends capital. The infrastructure layer generates free cash flow, grows dividends, and compounds shareholder returns. This is the AI cross-pollination dynamic: every dollar of capex from the model builders creates revenue for the infrastructure providers, who are publicly traded, FCF-positive, and governed by boards that demand capital discipline.

Landsman's hire is real. The need for an enterprise channel is real. But the conclusion that this hire changes OpenAI's fundamental investment profile doesn't follow from the data. It doesn't address the $14 billion burn. It doesn't shorten the path to 2029 profitability. And it certainly doesn't change the fact that OpenAI is a private company with no dividend, no free cash flow, and a valuation that requires flawless execution for the next four years to justify its price.

Of the publicly tradeable AI beneficiaries, I favor Broadcom for its custom chip pipeline with OpenAI, Google, and Meta, its $26.9 billion in free cash flow, and its consistent dividend growth. Nvidia deserves a place in the portfolio for its GPU dominance and $119 billion in annual free cash flow, though its valuation already reflects much of the AI thesis. Microsoft benefits structurally from the AI ecosystem regardless of whether OpenAI's profitability thesis works out.

Comparisons between OpenAI's revenue growth trajectory and the cash-generating power of the infrastructure layer are not only unjustifiable; in my opinion, they are irresponsible. Revenue growth without free cash flow is a promise. Free cash flow with a dividend is a delivery.

For investors who want AI exposure but refuse to carry the binary risk of a private company burning $14 billion a year, an overweight position in the infrastructure layer - Broadcom, Nvidia, Microsoft, and the data center operators they feed - makes sense. These companies are getting paid today to build the future that OpenAI is financing. The partnership hire doesn't change that. Nothing in OpenAI's current financial structure changes that. The smart allocation follows the cash, not the press release.

OpenAI's Partnership Hire Doesn't Solve Its Cash Problem