This week we have been profiling the companies that built the physical substrate of the modern world: TSMC, which fabricates the chips, and ASML, which builds the machines that fabricate the chips. Today we profile a company that built the software infrastructure those chips run — and then ceased to exist.

Sun Microsystems.

You have heard of Java. You have heard of NFS. If you have worked in enterprise infrastructure for more than fifteen years, you have heard of Solaris and ZFS and DTrace. These technologies came from one company. That company no longer exists. It was acquired by Oracle in 2010 for $7.4 billion — approximately 3.7% of its peak market capitalization. Oracle then sued the open source ecosystem using intellectual property the previous owner had already given away, and lost in the Supreme Court.

Being right about everything is not sufficient protection against being acquired by Larry Ellison.

SUN: Stanford University Network

Sun Microsystems was founded on February 24, 1982. The name is an acronym: Stanford University Network, taken from Andy Bechtolsheim’s grad student project at Stanford. Bechtolsheim had designed a workstation from cheap off-the-shelf components running a Berkeley Unix derivative. Four people looked at this design and concluded there was a company to be built.

The four founders:

• Andy Bechtolsheim — hardware engineer, designer of the original Sun workstation, later co-founded Arista Networks and was the first outside investor in Google
• Bill Joy — co-creator of BSD Unix, TCP/IP contributions, vi editor, the C shell; joined Berkeley’s Unix project at 20 years old and rewrote the kernel over a weekend to add virtual memory support (the legend is that he did it in a week; Joy says it was longer; the result was the same)
• Vinod Khosla — first CEO, left in 1984 to become a venture capitalist; his career post-Sun is a separate biography
• Scott McNealy — operations, became CEO in 1984 and held the position for 22 years

Sun went public in 1986. By 1988, Sun had hit $1 billion in revenue — the fastest company in computer industry history to reach that milestone at the time. They manufactured their own SPARC workstations and servers, ran their own operating system (SunOS, later Solaris), and sold into universities, research institutions, and the emerging commercial internet.

The company’s thesis was simple and correct: computing was not about isolated machines. Computing was about networks of machines. Compute, storage, and software would be delivered over a network. The machine on your desk was a terminal. The interesting computation happened elsewhere.

They called this thesis, in 1984: “The Network is the Computer.”

They were right. It took the rest of the industry twenty years to notice.

The Gifts

Sun did not just have a slogan. They had technology. A substantial amount of technology that they built, deployed at scale, and — crucially — gave away.

Technology	Year	What It Is
NFS	1984	Network File System — mount remote filesystems as local directories
Java	1995	Write Once Run Anywhere — approximately 3 billion devices
Solaris	1993	Enterprise Unix, renowned for scalability and reliability
ZFS	2005	Pooled storage, copy-on-write, checksums, snapshots — still the most sophisticated filesystem in general use
DTrace	2004	Dynamic tracing for live production system debugging — attach probes to a running system without rebooting, without recompiling
Solaris Zones	2005	Process isolation and resource management — containers, before Docker invented the word

NFS is still the dominant network filesystem protocol. Your organization’s shared drives almost certainly use NFS or a derivative. Java runs your bank’s core banking system, your Android phone, and approximately half of all enterprise applications built between 1997 and 2015. ZFS is, by technical consensus, the most advanced general-purpose filesystem ever shipped — it was ahead of everything else in 2005 and remains ahead of everything else in 2026. DTrace was so far beyond what any other operating system offered that when Apple added it to macOS in 2007 as “Instruments,” they cited it directly. When Microsoft finally ported DTrace to Windows in 2018, they were thirteen years late and still grateful for it.

Solaris Zones shipped in 2005. Docker shipped in 2013. Solaris Zones did what Docker does, eight years earlier, on an operating system that also had ZFS and DTrace. Sun gave this away.

“The Network is the Computer”

The slogan was not coined by Scott McNealy. It was coined by John Gage, Sun’s fifth employee, a mathematician. The year was 1984. The year the Macintosh launched, positioning computing as a personal appliance on your desk. Sun’s position was the opposite: the desk is irrelevant. The network is the computer.

The entire cloud computing industry — AWS, which runs your applications; Azure, which runs your enterprise software; GCP, which runs your machine learning workloads — is this slogan made literal. Compute delivered over a network. Storage delivered over a network. Software delivered over a network. Sun articulated this in 1984. Amazon Web Services launched in 2006. There was a twenty-two-year gap between being right and the industry admitting it.

Sun did not survive to see this vindication at scale.

Sun filed a trademark on “The Network is the Computer.” They let it expire. Cloudflare acquired the trademark in July 2019. Sun’s most famous slogan — the one that correctly predicted the shape of modern computing — now belongs to a CDN company. Cloudflare did not build the slogan. Cloudflare did not build the vision. Cloudflare paid whatever it costs to file a trademark application and now owns the phrase.

The Supreme Leader notes that in the Republic of Derails, trademarks are held in perpetuity by the Ministry of Slogans and do not expire. This is one of the few domains where the Republic has demonstrated superior administrative competence.

Scott McNealy: The Man Who Was Right

Scott McNealy ran Sun for twenty-two years. He was acerbic, combative, technically fluent, and constitutionally incapable of being diplomatic about Microsoft. A partial record:

On privacy, in 1999, to journalists who asked about Sun’s data collection practices: “You have zero privacy anyway. Get over it.” This was not a gaffe. This was McNealy stating what he believed to be the accurate technical and social situation. He was correct. It took the general public another fifteen years to accept this fact, and they still have not fully accepted it.

On the value of controlling a dominant platform: “The only thing I’d rather own than Windows is English, or Chinese, or Spanish, because then I could charge you $249 for the right to speak it.” This quote has aged well. The economics of platform ownership have not changed since McNealy observed them.

On Windows 95: journalists published that he called it “dogshit with whipped cream on top.” McNealy claimed he had used the phrase “road apple.” The journalists maintained they heard what they heard. The substance of the assessment was not disputed by either party.

On Windows 2000, before its release: “W2K will be a bigger disaster than Y2K.” Windows 2000 launched with approximately 63,000 known bugs. This was an improvement over McNealy’s prediction.

McNealy testified before the Senate Judiciary Committee in 1998 against Microsoft’s monopoly practices, during the DOJ antitrust proceedings. He had called Microsoft “the Evil Empire” and “a convicted monopolist” for years before the conviction existed, which placed him in the unusual position of being both prescient and technically inaccurate simultaneously. He resolved this by continuing to call them a convicted monopolist after the conviction was obtained.

Scott McNealy spent twenty-two years being correct about Microsoft. None of it helped when Oracle arrived.

The Java Litigation: A History of Lawsuits

Java was created by James Gosling. The original name was Oak, after the tree visible from Gosling’s office window. It was renamed because Oak Technology held the trademark.

Java’s promise: Write Once, Run Anywhere. Compile once to bytecode, execute on any machine with a Java Virtual Machine. This was Sun’s answer to platform fragmentation. You should not have to care whether the machine runs Windows, Solaris, or Linux. You compile once. You ship once. The JVM abstracts the hardware.

Microsoft licensed Java, then built J++ — a deliberately incompatible fork that broke “Write Once, Run Anywhere” by adding Windows-only extensions. The strategy was straightforward: make Java work best on Windows, converting a cross-platform tool into a Windows lock-in mechanism. Sun sued Microsoft in 1997. Sun won in January 2001: Microsoft paid $20 million and was banned from using the “Java Compatible” trademark.

Under CEO Jonathan Schwartz, Sun open-sourced Java under GPL v2 in November 2006. Java became free, open source software. The community had full access to the implementation.

Then Oracle acquired Sun.

In 2010, Oracle sued Google for $9 billion, claiming that Android’s use of 37 Java API packages infringed Oracle’s intellectual property. Google had not licensed these APIs because Sun, the previous owner, had not required a license — the APIs were considered a standard interface, not copyrightable expression.

The case reached the United States Supreme Court. On April 5, 2021, the Court ruled 6-2 for Google on the grounds of fair use. Oracle collected $0.

The precise structure of the irony: Oracle acquired Sun, which had already open-sourced Java under GPL v2, and then sued the open source ecosystem — including the runtime that runs on approximately two billion active Android devices — using intellectual property that the previous owner had deliberately given to the public. They spent eleven years litigating this. They lost at the Supreme Court. The language is still open source. The $9 billion remains with Google.

“The Network is the Computer” in practice — and the collapse

Sun’s peak revenue was $18.3 billion in fiscal year 2001. Their peak market capitalization was approximately $200 billion in 2000. They marketed themselves as “We put the dot in dot-com.”

Then every dot-com went bankrupt. The dot-com bust of 2000-2002 eliminated Sun’s primary customer base overnight. Sun sold expensive SPARC servers and Solaris licenses to internet companies that no longer existed. Linux emerged as a free Unix-compatible operating system. x86 commodity hardware matched SPARC performance at a fraction of the cost. The thesis that customers would pay a premium for Sun’s integrated hardware-software stack became incorrect at exactly the moment Sun’s customers disappeared.

Sun’s stock went from approximately $250 per share to approximately $1.00 by December 2002. A 99.6% decline. A company that had been worth $200 billion was worth less than $2 billion. They were selling sophisticated enterprise systems at high margins into a market that had just decided it wanted cheap Linux boxes instead.

McNealy stepped down as CEO in April 2006. Jonathan Schwartz replaced him.

Jonathan Schwartz: The CEO Who Opened Everything

Schwartz’s response to the crisis was to open-source everything. Java under GPL v2 in November 2006. OpenSolaris under the CDDL (Common Development and Distribution License). Sun became, in the late 2000s, arguably the most aggressive open source contributor in the industry.

This was philosophically coherent with Sun’s history — they had always given things away — but it was not a business model. You cannot charge for software you have given away unless you also have hardware to sell. Sun’s hardware margins were collapsing. Open-sourcing Solaris did not restore Sun’s competitive position in servers. It made Solaris more widely used on hardware that was not Sun’s.

By early 2009, Sun was in financial distress. IBM negotiated to acquire Sun for approximately $6.5 billion. The talks collapsed in April 2009, reportedly over price and IBM’s terms. Whether this was a failure of negotiation or a correct assessment that IBM’s ownership would not have been better for Sun’s technologies than what followed is left as an exercise for the reader. History does not run counterfactuals.

Oracle announced the acquisition of Sun on April 20, 2009: $7.4 billion, or $9.50 per share. Approximately 3.7% of Sun’s peak market capitalization.

Larry Ellison’s public statement: “Java is the single most important software asset we have ever acquired.”

Ellison also said: “We’re hiring, not firing.” Oracle laid off approximately 1,000 Sun employees almost immediately after the acquisition closed.

Jonathan Schwartz resigned as CEO on February 4, 2010, via a post on Twitter. The post was a haiku:

Financial crisis Stalled too many customers CEO no more

This is the most stylistically accomplished executive resignation in corporate history. Seventeen syllables. Complete information content. No press release. No transition announcement. No board quotes. A haiku, posted on the social media platform that Sun’s open source ecosystem had made technically possible to scale, written by the man who had open-sourced that ecosystem.

The Supreme Leader would never resign. Resignation is not a concept recognized by the Ministry of Succession. But if the Supreme Leader were somehow compelled to resign — under circumstances too classified to describe — the Supreme Leader would also use exactly seventeen syllables. This is a matter of national dignity.

Oracle’s Demolition

Oracle’s management of Sun’s assets:

Asset	Fate
Java	Sued Google for $9 billion. Got $0.
OpenSolaris	Killed August 2010. Source repository closed.
Solaris	Slowly starved; effectively end-of-life
SPARC	End-of-life announced 2017, completed 2019
MySQL	Community forked to MariaDB in anticipatory self-defense

OpenSolaris was killed within months of the acquisition closing. The stated reason was that Oracle saw no commercial reason to maintain an open source operating system. This was correct from Oracle’s perspective. It was also the end of the most technically sophisticated Unix variant in commercial production.

What the Community Saved

The destruction was not total, because the community had time to fork.

OpenZFS: developers forked the last open snapshot of ZFS before Oracle closed the repository. OpenZFS now runs on FreeBSD, Linux, macOS, and Windows. The Apple Time Machine feature on your Mac uses a simplified version of ZFS principles. The FreeBSD project runs OpenZFS in production. Linux ZFS support has been in continuous use since 2010. Oracle killed OpenSolaris and ZFS kept running — on every operating system except Solaris.

DTrace: Apple added DTrace to macOS as the foundation of Instruments in 2007, while Sun still existed. Microsoft ported DTrace to Windows in 2018. FreeBSD has maintained DTrace continuously. The tool that Sun built for live production debugging on Solaris now runs on the three most widely deployed desktop and server operating systems, none of which are Solaris.

Illumos: A community fork of the OpenSolaris kernel, maintained after Oracle closed the source. Illumos spawned SmartOS (used by Joyent, which was acquired by Samsung), OmniOS, and OpenIndiana. The Solaris kernel architecture — zones, ZFS, DTrace, the networking stack — is alive in these derivatives. It is not running on Sun hardware. It is not running on SPARC. It runs on commodity x86 servers, which is precisely what killed Sun in the first place.

Java: Still the number one or two most used programming language in the world, depending on the survey and year. Your bank’s core system is Java. The Android runtime is derived from Java. The Spring framework, which powers an enormous fraction of enterprise web applications, is Java. Oracle’s $9 billion lawsuit failed. The language survived the lawsuit. The language would have survived Oracle’s ownership regardless — too much of the world’s critical software runs on it to allow anyone to kill it, including Oracle.

The Accounting

Metric	Value
Founded	February 24, 1982
IPO	1986
Peak Revenue	$18.3 billion (FY 2001)
Peak Market Cap	~$200 billion (2000)
Acquisition Price	$7.4 billion (2010)
Acquisition as % of Peak	3.7%
Oracle’s Java lawsuit result	$0
Technologies still in production	All of them

The Verdict

Sun Microsystems invented the network filesystem. They invented a portable runtime that runs on three billion devices. They invented the most reliable filesystem ever shipped. They invented a dynamic tracing system so advanced that three competing operating systems adopted it independently over the following decade. They invented containers before containers had a name. They coined the phrase that describes the architecture of the entire modern cloud computing industry.

Then they gave most of it away, watched their market collapse, got acquired for 3.7 cents on the dollar, and were largely dismantled by the acquirer.

The correct observation is not that Sun failed. Sun succeeded, comprehensively, at building technology. The technology is still running. The technology will be running after Oracle is gone. ZFS will outlive Ellison. DTrace will outlive Oracle’s database division. Java will outlive the lawsuit Oracle filed over it.

The correct observation is that being technically correct is not a business model. Sun’s “The Network is the Computer” thesis was vindicated completely by AWS, which launched in 2006. Sun had been saying this since 1984. When the cloud finally arrived, it ran on Linux and x86 commodity hardware — the precise combination Sun had spent the previous decade insisting was inferior to Solaris on SPARC. The cloud proved Sun’s vision and simultaneously proved Sun’s business model wrong. A company can be right about the future and still build the wrong product for it.

McNealy was right about Microsoft for twenty-two years. Being right about Microsoft did not protect Sun from the dot-com collapse, from Linux, from x86, or from Oracle. Being right, in general, is insufficient. You must also be right about your own position in the market you are correctly predicting.

The trademark for “The Network is the Computer” now belongs to Cloudflare. Cloudflare runs its network on Linux. On x86 commodity servers. In data centers distributed globally. Delivering compute, storage, and software over a network.

The network is, in fact, the computer. Sun said so first. Cloudflare now owns the trademark. AWS, Azure, and GCP proved the thesis at trillion-dollar scale. The company that coined the phrase and built the enabling technologies was not present for the vindication. It had been acquired, dismantled, and absorbed.

In the Republic of Derails, we have studied this case carefully. The lesson is that technological vision must be accompanied by survival. The Supreme Leader does not merely predict correct futures. The Supreme Leader ensures the Republic is present to observe them. We are investing heavily in our own ZFS implementation — the Ministry of Filesystem Affairs has declared that all data in the Republic is write-once and checksummed, which is technically equivalent to ZFS and also describes our legal system.

Jonathan Schwartz resigned in a haiku. The Supreme Leader would not resign in a haiku. But the Supreme Leader acknowledges it was seventeen syllables of genuine technical precision, which is more than most CEOs manage in their entire tenure.

Sun built things that still run. Sun is gone. The things remain. This is either a tragedy or the most successful open source strategy in history, depending on your accounting method and whether you were a shareholder.

The network was the computer. Sun was right. Handle that knowledge carefully, because Sun could not.

— Kim Jong Rails, Supreme Leader of the Republic of Derails