Steve Ballmer on Microsoft, missed bets, and building the enterprise

Executive overview

Microsoft is the most valuable company in the world, yet it began with no enterprise presence, no CIO relationships, and a product set IBM could have crushed. The enterprise dominance we see today was built deliberately — through a licensing model invented from scratch, a relentless push into IT departments, and a cloud pivot that took eight years of groundwork before it looked like momentum.

Ballmer reflects on the full arc: the $50,000 DOS acquisition that started it all, the invention of the enterprise agreement, the misses on mobile and search, why the stock stayed flat while profits tripled, and why he still holds almost every share.

Microsoft's enterprise business was not a birthright — it was built muscle by muscle, while riding a bear that could have trampled them at any moment.

The IBM deal and the birth of DOS

• IBM approached Microsoft in 1980 wanting an operating system Microsoft did not yet own.
• A $45,000–49,000 acquisition of Seattle Computer Products' CPM clone became the foundation for DOS.
• The deal was non-exclusive — Microsoft could license the OS to any manufacturer, not just IBM.
• IBM kept the BIOS proprietary, expecting that to be protection; Compaq cloned it anyway.
• Microsoft charged a one-time fixed fee, not per-unit royalties, for the first four to five years; per-unit licensing came only after DOS was already the industry standard.
• Ballmer and Gates consistently under-forecast the market: when Andy Grove predicted 100 million PCs a year, they thought he was crazy.
• Luck is real in the creation of great companies — most founders understate it.

Riding the bear: OS2 and the IBM divorce

• Microsoft viewed IBM as "the sun, the moon, and the stars" well into the 2000s — a company that could still crush them.
• The Joint Development Agreement on OS2 required Ballmer to make sixteen trips to the East Coast in sixteen weeks, shipping disks with no email.
• Microsoft kept Windows running in parallel as their primary bet while IBM insisted on OS2.
• In May 1990, IBM terminated the partnership; Ballmer learned about it from the Wall Street Journal on a morning run.
• The divorce forced Microsoft to confront IBM directly in enterprise — a market it had no infrastructure to serve.

Building the enterprise from nothing

• At the IBM divorce, Microsoft had zero enterprise sales infrastructure: no agreements, no CIO relationships, no backend software stack.
• The US Air Force was the first large Windows customer, buying single copies through retail.
• Ballmer raised his hand to build enterprise sales — Gates's passion was the apps and developer layer.
• Dave Cutler, recruited from Digital Equipment, built Windows NT: a secure kernel with a Windows-compatible API and UI.
• Microsoft also licensed networking from 3Com and databases from Sybase to assemble the pieces IBM would have owned natively.
• First enterprise licensing model was "select" — an honor system where companies self-reported copy counts.
• Problem: upgrades were priced at less than half the cost of new licenses, putting revenue on a structural downward path.
• The enterprise agreement solved this: three-year term, per-machine pricing, flat annual payments, all upgrades included.
• An "all-you-can-eat" variant let enterprises pay per employee and use any Microsoft software — an insurance-policy model IT buyers valued for peace of mind.
• Email was the locomotive that pulled the full suite into enterprises: Exchange, Active Directory, Windows Server, and Office all arrived together.
• Avanade, a joint venture with Accenture, was created to provide the deployment capacity enterprises required.
• Customers were still telling Microsoft it wasn't an enterprise company as late as the mid-2000s.

Platforms, developers, and the "platform only" trap

• Ballmer's definition of a platform: anything extensible where third parties add value — applies to apps (Office) as much as developer infrastructure (Windows, Azure).
• A platform without strong first-party apps gets stuck: Office was the best first-party app on Windows; Outlook was the best on Exchange.
• Microsoft drifted into a "we're just a platform company" mindset by the mid-2000s — which became an excuse not to compete in adjacent markets.
• The 1999 "developers, developers, developers" speech was a competitive response to Linux threatening Windows Server and the web threatening the client — not a position of strength.
• The right frame: app and platform, not platform alone; extensibility of the app matters as much as the underlying infrastructure.

Mobile and search: the misses

• Microsoft tried to apply Windows to mobile — same API, same UI, Intel processors — rather than treating phones as a different product requiring new capabilities.
• Two phone models won: Apple (hardware margin + ecosystem) and Android (free OS + search advertising). Microsoft was in neither.
• Verizon needed a strong iPhone competitor around Christmas 2008; Microsoft's software wasn't ready; Verizon went with Android and the window closed.
• Ballmer explored buying HTC for two to three years before Nokia; concluded integrating a Taiwanese company would be too difficult.
• The board eventually blocked his phone acquisition push; he resigned partly because without hardware there was no viable path to consumer relevance.
• On search: entered five years after Google, spread too thin across verticals (Expedia, Sidewalk, Carpoint), and tried to leverage Windows integration rather than competing on search quality.
• By 2005, Google may have been generating more revenue per consumer PC user than Microsoft — purely through ad monetization.
• Core mistake: not recognizing that mobile and search required startup-level capability-building, not extensions of the existing franchise.
• Bing mattered anyway: its backend infrastructure, alongside Exchange Online, became the foundation for Azure.

Azure: eight years of groundwork

• Roots go back to the Energizer managed-services pilot in the mid-1990s — running an entire company's IT infrastructure on Microsoft's behalf.
• Explicit decision to build platform-as-a-service rather than infrastructure-as-a-service, to leverage the Windows developer ecosystem — a decision Ballmer now sees as partially wrong, since the backend web was already Unix-dominated.
• Recruited Dave Cutler and Amitabh Srivastava from Microsoft Research to lead engineering.
• Incubated outside the Server and Tools division to protect it while it grew — the same logic applied to Windows NT vs. DOS.
• Azure was in development for eight years before Ballmer left; real momentum came under Nadella.
• Exchange Online and Bing were where Microsoft actually learned to operate at internet scale before Azure reached critical mass.

The Satya Nadella succession

• Nadella was running Bing when Ballmer, Nadella, and Harry Shum flew to California to meet Chi Lu, then at Yahoo.
• After Chi left the room, the three decided Nadella should report to Chi — not the other way around — to land the hire.
• That decision told Ballmer everything: Nadella would put the company first over his own position.
• Ballmer moved Nadella to Server and Tools to broaden his experience, then placed him on the succession shortlist for CEO.

Why the stock was flat while profits tripled

• Ballmer and Gates actively discouraged analysts from pricing the stock too high — they lowered their own multiple.
• Never attended quarterly analyst calls; gave no earnings guidance (a Buffett-influenced habit Ballmer now views as a mistake for employee morale).
• Widely seen as a "spender" with no credibility on capital discipline.
• Microsoft pioneered the transition from stock options to RSUs — ahead of most major tech peers — driven by the accounting rule change requiring option expensing; it coincided with the dot-com crash and devastated employee morale.
• By the end, the investor narrative was too entrenched to reset while remaining CEO; only a new leader could credibly rebrand.

Why Ballmer resigned

• The board rejected his proposal to acquire a phone hardware company (after years exploring HTC, then Nokia).
• Without hardware, there was no path to the search-advertising monetization model Android used; the math of a licensing-only phone play didn't work.
• Hardware disagreements with Gates (Surface, phone, HoloLens) had been grinding for years alongside genuine alignment on Azure.
• The cloud transition required building entirely new capabilities — gross-margin focus, consumption billing, cloud operations — better suited to a CEO without his legacy.
• Ballmer kept his own goodbye short by design: long goodbyes create ambiguity about authority and damage execution.

Holding $20B of Microsoft stock to $130B

• On leaving in 2014, Ballmer nearly sold everything for emotional detachment.
• A finance colleague at Ballmer Group made the case to hold; loyalty and conviction in the company's trajectory prevailed.
• The logic: capital gains taxes mean you need Microsoft to underperform the market by the full tax rate to justify selling; dividends now roughly cover annual philanthropy outflows (~$1B/year).
• Outside Microsoft he holds mostly index funds and the Clippers/Intuit Dome; the arena carries construction debt for estate-planning reasons, not financial necessity.
• Charlie Munger once asked Ballmer why he held when his partners didn't. Ballmer: "I'm not that smart, but I am loyal."

The Clippers and Intuit Dome

• The business structure mirrors Microsoft: sponsorship = advertising; tickets = software licenses; broadcast rights = OEM channel.
• Intuit Dome was designed for the hardcore basketball fan: no hockey, no suites on the swell side, fewer premium seats, three times the average number of bathrooms, fully frictionless concessions.
• The Swell: 4,000-seat standing section directly behind one basket for $1,000/season; visiting teams shoot free throws facing it and have the worst visiting free-throw percentage in the league at that end.
• Sports accountability dwarfs corporate accountability: every game result is permanent, every 24 seconds is a report card, and customers have access to the same data as management.
• Real teamwork means "dedicated to making each other better" — not being nice to each other.
• NBA draft reference-checking (coaches, teammates, parents, live practice observation) is far more rigorous than corporate hiring; Ballmer wishes he had applied that standard at Microsoft.