Microsoft Volume II: From Browser Wars to Azure (1995–2014)

Executive overview

From 1995 to 2014, Microsoft grew revenue from $6B to $80B while simultaneously being dragged through an existential antitrust battle, losing mobile entirely, and failing in nearly every consumer category. Yet it ended the period as the foundation for one of the greatest business transformations in tech history.

The standard narrative — Microsoft won, then lost, then Satya came and won again — misses the real story. The "lost decade" under Steve Ballmer was when Microsoft quietly built Azure, cracked enterprise software distribution, and preserved the talent base that would eventually make it the world's most valuable company.

The internet and the browser wars (1993–1998)

• Microsoft almost missed the internet entirely — its early bets were on interactive television ("CableSoft") and walled-garden services (MSN/Project Marvel)
• A 1994 internal memo from engineer Jay Allard coined "embrace and extend" and correctly predicted the web browser as the next killer platform — months before Netscape existed
• Stephen Sinofsky's "Cornell is Wired" memo after being snowed in on campus convinced Bill Gates the internet was an exponential phenomenon, not a niche academic tool
• Gates published The Road Ahead in November 1995 still focused on the "information superhighway"; the paperback edition replaced every instance with "the internet"
• Microsoft licensed the Mosaic browser code for $2M and shipped Internet Explorer; Gates made it free and bundled with Windows 95 on December 7, 1995 — Netscape stock fell a third that day and never recovered
• By 2000, Internet Explorer held ~100% browser market share; Netscape was acquired by AOL in 1998 for $4B in stock
• The 1997 Apple bailout deal — $150M investment, five-year Office-for-Mac commitment — also locked Internet Explorer as the default Mac browser, displacing Netscape from its last refuge

The DOJ antitrust case (1998–2001)

• The case originated with an FTC investigation in 1990 that deadlocked 2–2 in 1993; the DOJ immediately picked it up — effectively prosecuting Microsoft twice for the same conduct
• The central legal question was whether Internet Explorer was a "product" (which couldn't be tied to Windows under the 1994 consent decree) or a "feature" (which could be integrated)
• Prosecutor David Boies deployed videotaped Bill Gates depositions strategically at trial — Gates appeared evasive and combative; the clips shaped both the ruling and public opinion
• In November 1999, Judge Jackson found Microsoft was a monopoly; in June 2000, he ordered the company broken up into two separate entities — an OS company and an applications company; Gates and Ballmer were barred from working at the same one
• Microsoft's market cap fell from $600B to $270B in 12 months, coinciding with the dot-com bust and the CEO transition
• The breakup order was ultimately reversed on appeal; Microsoft settled with a consent decree that amounted to a slap on the wrist — but five years of litigation drained leadership focus and, per multiple insiders, caused what one described as a "mental breakdown for the whole company"
• Brad Smith, promoted to general counsel by Ballmer, presented a single-slide deck to the board: "It's time to make peace." He then systematically settled every outstanding government case globally

Steve Ballmer's tenure and the enterprise machine

• Ballmer's three priorities on taking over: hold the company together emotionally, clean up antitrust, keep growing — he succeeded on all three
• The enterprise agreement (EA) model — multi-year volume licensing deals with large corporations — became the engine of Microsoft's revenue growth and its greatest structural advantage
• EAs aligned Microsoft with IT administrators rather than end users: IT valued backward compatibility, security controls, and stability; they were willing to suppress new features to avoid retraining costs
• One Office executive told researchers: "I always thought we could stop bundling new features for ten years and it would be fine. No one would notice. People would probably pay more for it."
• The EA model created a flywheel: lock-in at the IT buyer level, guaranteed upgrade cycles, and an army of resellers and consulting partners (Accenture et al.) whose livelihoods depended on Microsoft staying the standard
• Microsoft grew enterprise revenue through nearly every year of the Ballmer era, including the 2008 recession — the financial statements looked completely different from the consumer narrative

Consumer failures: Vista, Zune, Windows Phone, and Windows 8

• Vista (code-named Longhorn) was intended as a radical architectural rebuild; after years of scope creep, it shipped in 2007 stripped of its key innovations and was widely panned for performance and compatibility failures
• The Zune launched in 2006 against the iPod and never gained traction; Microsoft cancelled it in 2011
• Windows Mobile was built for enterprise handsets and operated like an ingredient product — OEMs could modify the OS code freely, resulting in inconsistent experiences and no platform coherence
• The iPhone's 2007 launch changed the frame entirely: it kicked off "shadow IT" and the bring-your-own-device movement, eroding the enterprise lock-in Microsoft had spent a decade building
• Microsoft missed a critical window because Android undercut any Windows Phone licensing model — Google gave Android away for less than free, subsidising carrier adoption through marketing dollars; Microsoft's business model required charging for the OS
• The $47B Yahoo bid in 2008 collapsed at a Boeing Field meeting when Ballmer and Gates concluded the Yahoo founders didn't genuinely want to merge; Microsoft never became a meaningful search player
• Bing launched in 2009 — eleven years after Google — and captured only modest share; search's winner-take-most economics (more queries → better data → more advertisers → higher prices) meant second place was structurally unprofitable
• Windows 8 attempted to unify tablet and desktop into a single touch-first OS but shipped a confused product: the Metro tile interface was designed for tablets but forced onto desktops, OEMs weren't ready for the touch hardware requirements, and the developer ecosystem never materialised
• Ben Gilbert, a Microsoft employee from 2011–2014, recounted that Office for iPad was cancelled mid-development in 2013 to protect the Surface marketing message; Satya Nadella shipped it as one of his first acts as CEO in 2014

Xbox: the right market, the wrong returns

• Xbox launched in 2001 and Xbox Live by 2002, pioneering the subscription-based online gaming service model that would eventually have 40M subscribers
• The Entertainment & Devices division ran at or near breakeven for most of its existence: $1.4B operating loss on $4B revenue in 2006; modest profit on $8B revenue by 2008–2010
• The real Xbox dividend was organisational: the team built Microsoft's first large-scale always-on internet service, providing the DNA and infrastructure experience that directly informed Azure
• Ballmer's own quote when asked about failed acquisitions: "We only lost money" — pointing to talent and focus as the truly scarce resources, not capital

The secret origin of Azure (2005–2014)

• In 2005, Ray Ozzie — recruited as Gates's successor as chief software architect — wrote the "Internet Services Disruption" memo outlining the cloud services model before anyone used the term cloud
• Ozzie recruited Dave Cutler (architect of Windows NT) to lead a secret incubation project code-named Red Dog, deliberately kept outside the Windows and server-and-tools divisions to avoid internal organ rejection
• The organisational logic: the existing server business sold on-premises Windows Server through OEMs and Accenture-led implementations; a cloud service would directly undercut that go-to-market and every partner's livelihood
• Azure launched in 2008 as a platform-as-a-service product running Windows Server — deliberately not infrastructure-as-a-service like AWS, and refusing to support Linux or open source
• In 2010, Ballmer gave a public speech at the University of Washington declaring "we are betting the company on cloud and Azure," replaced server-and-tools division head Bob Muglia, and installed Satya Nadella — then running Bing — to lead the transformation
• Nadella's appointment was as much about executive development as product strategy: the Azure role was designed to give him the profile to become CEO
• The talent pipeline for Azure drew from across Microsoft: Bing supplied distributed-systems expertise; Xbox Live provided always-on service operations at scale; Hotmail contributed high-volume web application experience; the Windows Server group provided hypervisor and OS kernel depth
• By fiscal 2023, the Intelligent Cloud segment (Azure plus SQL Server and Windows Server) generated $88B in revenue — the company's largest and fastest-growing division
• The widely-held narrative that Satya invented Azure is wrong. Azure was conceived by Ray Ozzie, built by Dave Cutler's team, championed through seven years of internal resistance by Steve Ballmer, and handed to Nadella as a going concern. Satya deserves credit for execution and cultural reset — not vision.

Nokia acquisition and the end of Ballmer's tenure

• Nokia adopted Windows Phone as its primary OS in 2011; by 2013 it was clear Android had won and Nokia was threatening to switch
• Ballmer proposed the $7.5B Nokia mobile-unit acquisition; a majority of division heads, including Nadella, opposed it; the board pushed back
• On August 23, 2013, Ballmer announced his retirement; ten days later, Microsoft agreed to buy Nokia — the question of who actually authorised the deal remains genuinely unclear
• Nokia's device business was eventually written down almost entirely; the acquisition is widely regarded as a costly final bet on a lost mobile war
• February 4, 2014: Nadella becomes CEO; Gates steps down as chairman; Thompson takes the chair; an explicit generational handover closes the Ballmer era

Seven powers analysis

• Scale economies: Microsoft's single greatest enduring advantage — any R&D or infrastructure investment is amortised over an enormous user base, translating directly to cloud economics
• Switching costs: extremely high in enterprise through EA lock-in, backward compatibility guarantees, and ecosystem depth; much lower in consumer
• Network economies: strong in developer platform and enterprise; nearly absent in consumer and mobile
• Counter-positioning: lost as an incumbent — Google repeatedly counter-positioned Microsoft by giving away free alternatives to things Microsoft charged for (Android, Gmail, Google Docs)
• Process power: eroded significantly during the Vista and Windows 8 cycles; partially rebuilt under Nadella
• Branding: strong in enterprise, became a liability in consumer during the 2005–2013 period
• Cornered resource: Azure's early enterprise relationships and hybrid cloud positioning created a temporary moat versus AWS

Playbook themes

• Embrace and extend as a durable strategy: Jay Allard's 1994 formulation — find an exponential phenomenon with demonstrated product-market fit, integrate it into the platform, monetise through existing channels — worked for browsers and later for cloud
• Positive-sum leadership: Gates and Ballmer were far more valuable together than either alone; Gates provided technical vision and willingness to change direction on new data; Ballmer provided execution, alignment, and the emotional rock the company needed during the antitrust crisis
• Partner orientation is both moat and trap: Microsoft's ecosystem of resellers, ISVs, and consulting partners created enormous lock-in but made it nearly impossible to reset for a new era — Apple, with fewer external dependencies, could do a full architectural reset when Jobs returned; Microsoft could not
• Narrative failure as strategic risk: Microsoft's stock price stagnated for a decade despite tripling revenue and profits; the company failed to tell its own story — "we invent and wander, we make bold bets" in the way Amazon did. The self-reinforcing "Microsoft sucks" narrative depressed talent attraction and market cap simultaneously

The real verdict on the "lost decade"

• Microsoft grew revenue from $6B (1995) to $80B (2014) and maintained profitability through almost every year
• Of the peer set from 2000 — Yahoo, AOL, Nortel, HP, the media conglomerates — virtually none survived intact; Microsoft, Oracle, Dell, and a bailed-out Apple are the exceptions
• Consumer products languished and mobile was entirely missed, but the enterprise machine never stopped compounding
• Azure was being built for seven years before Satya became CEO
• The "lost decade" framing is the company's greatest communication failure — and arguably its most consequential one, given how much talent and capital allocation a sustained narrative shapes