Softwar: An Intimate Portrait of Larry Ellison and Oracle $ORCL (Fintwit Book Club July 2025)
Summary
Both Andrew Walker and Byrne Hobart say that, had they read the book in 2003, they would have gone long QQQ and short Oracle, although the portrait’s red flags obscured how much Larry Ellison got right. Oracle looked chronically vulnerable: quarter-end sales, channel stuffing, related-party dealings, executive churn, and a CEO preoccupied with boat racing. Yet Hobart could not decide whether it was “a good company selling products in a hard category” or a company with adequate technology that was exceptionally good at selling and habitually overpromised.
An ERP migration is less a server upgrade than a “partial brain transplant” performed while the company stays alive. Oracle’s own implementation effectively shut the company down for 10 days over Christmas; Andrew recalled a GE Power project—possibly in Hungary—where bolted-on procedures could have produced deadly consequences in a live plant. The stories made the perennial “short the ERP transition” trade intelligible—and gave context to Lamb Weston’s later software-driven earnings blowup.
Oracle’s obsession with quarterly earnings and its stock price may have been product strategy as much as financial vanity. For an enterprise buyer, a recognizable vendor with positive media coverage, a rising stock chart, and apparent financial health supplies social proof that the company will survive. Oracle’s Financial Times advertisements reinforced that message, while the stock also determined whether engineers said, “I just bought a house for cash,” or felt poorer after holding their options.
Ellison repeatedly condemned traits he also displayed: salesmanship, exaggeration, and rule-bending. He called Oracle an engineers’ company while the book mostly shows him selling to CEOs, dismissed salespeople while personally deflecting customers who had spent tens of millions with no return, and attacked competitors’ dishonest benchmarks while Oracle advertised extraordinary ones. Hobart’s summary was that Ellison “hates people who remind him of his own worst traits.”
Ellison’s durable insight was that centralized data, browsers, cheap devices, and standardized software would reorganize how people behave. His roughly $500, mobile-ish browser device resembles the eventual smartphone, while his arrogant instruction—adapt business processes to Oracle rather than customize Oracle—anticipated software putting users “on tracks that are written in code.” That vision mattered more than whether he personally appeared in the book debugging shell scripts.
Oracle’s combative culture nevertheless became a remarkable executive factory. Marc Benioff runs Salesforce, Tom Siebel ran Siebel Systems, and other Oracle executives went on to lead major technology companies. Oracle later bought Siebel for roughly $6 billion after years of personal warfare.
Safra Catz’s eventual succession shows why the next CEO may need to be unlike the founder. The book’s author dismisses her as too loyal and ill-suited to succeed Ellison, while Ellison describes an ideal successor who sounds nothing like her; she nevertheless rises through a position where “basically all communication was going through Catz.” Hobart’s broader lesson: a great incumbent solves the problems matching their strengths, leaving a different set of problems—and therefore a different leadership profile—for the successor.
Deep dive
1. The obvious 2003 short case missed Ellison’s technological foresight
Andrew’s opening test was deliberately unforgiving: after reading the 2003 portrait, would either investor buy Oracle or short it against QQQ? Both ultimately chose relative underperformance, though Hobart emphasized how uncertain the book left him.
Hobart’s honest non-answer: Oracle might be “a good company selling products in a hard category,” or merely brilliant at selling acceptable technology while habitually overpromising. Either interpretation could produce dissatisfied customers; only one implied eventual corporate failure.
His likely 2003 mistake would have been dismissing distributed systems, commodity hardware, and Unix written by “crazy amateurs” as unsuitable for serious work compared with Sun. With hindsight, Ellison’s architecture was much closer to the eventual direction of computing.
Ellison also described a roughly $500, small, mobile-ish device used constantly through a browser. He thought laptop, not smartphone, but Hobart saw the modern model: most nongaming apps are effectively “a topic-specific browser” presenting the same information and capabilities as their websites.
2. ERP transitions resemble live brain surgery
Andrew had known the investor trope—“company implementing an ERP transition, short the stock”—without fully understanding it. Hobart supplied the mechanism: it is “kind of like a partial brain transplant,” replacing part of the brain stem while everything else must keep functioning.
Oracle’s migration to its own product left the company essentially unable to operate for 10 days. Management was grateful it had scheduled the change over Christmas, an extraordinary failure mode for a technology company implementing its own system.
Andrew recalled a GE Power example, possibly in Hungary, where the team tried attaching local systems and procedures to Oracle before concluding that, if the situation had been live, people in the plant might have died. The painful resolution was to change operating procedures to fit the software.
The contemporary specimen was Lamb Weston: software-focused growth funds had discovered a potato stock through 13Fs, only to risk losing money on potatoes after an ERP-related earnings blowup. The technology changed; the implementation hazard remained “completely timeless.”
3. Oracle read like a short seller’s governance checklist
Andrew’s dossier began with sales clustering on the final day of each quarter and persistent channel-stuffing concerns from the early 1990s onward. Oracle had already nearly died once, making another blowup seem possible, though not predictable in timing.
The related-party pattern appeared at inception: as Andrew retold it, Ellison directed work from his employer toward the company he had founded, despite doing little of the programming, and took most of the equity. Later, companies in which he had invested would partner with Oracle or become acquisition candidates.
Ellison devoted immense attention to boat racing, Oracle sponsored it, analysts questioned how much time he would take away, and he nearly died doing it. Andrew viewed the distracted CEO, related-party dealings, quarter-end pressure, and channel-stuffing concerns as a formidable cluster of red flags.
Executive treatment added another: Ellison repeatedly elevated a right-hand lieutenant, soured on him, and fired him shortly before options vested. His rationale—an assistant might receive only two weeks’ pay, so why should an executive deserve millions—was coherent in isolation but disturbing as a repeated pattern.
4. Ellison was the salesman he claimed not to understand
Hobart’s central contradiction: Ellison described Oracle as an engineers’ company, yet the narrative never shows him consulting a man page, debugging a shell script, or proving anything. It repeatedly shows him pitching CEOs and escaping complaints that customers had spent tens of millions without receiving a return.
Ellison complained that technology companies lied and manipulated benchmarks, but Hobart dryly observed that Oracle would command a much larger market capitalization “if every benchmark that I’ve ever seen in an Oracle ad were in fact true.”
His management reversals followed the same pattern. One celebrated salesperson received a mandate to “sell, sell,” then drew Ellison’s anger for doing exactly that with inadequate margins—even while Ellison argued that becoming the software standard made every incremental point of market share more valuable.
The book’s footnote device preserved the pettiness beautifully. After someone remembered Ellison greeting him in “a pink tank top” with carrot juice, Ellison corrected the record: “I have never owned nor would I ever wear a pink tank top. That is very important.” Hobart called the apparatus a book-length version of “Wasn’t Me.”
5. Oracle’s public image was embedded in the product
Andrew had rarely encountered a business history so fixated on earnings reactions and share price outside Enron. After one miss, Oracle even bought Financial Times advertisements explaining why its business remained strong and Wall Street was wrong.
Hobart’s explanation was social proof: if an enterprise customer has heard of the vendor, seen it on CNBC, and watched its stock chart rise, the customer feels more comfortable that the company will remain viable. “The bigger the contract and the bigger the company,” the more that credibility matters.
Share performance also shaped technical talent. Salespeople received commissions immediately and could blame themselves when pay declined; engineers holding options could shift from stories of buying houses for cash to feeling poorer because they owned and bought more stock.
Sell-side analysts then carried more genuine informational weight, while Gartner and Forrester could redirect budgets. Their modern analogue may be the venture-capital logo: seeing Sequoia on a startup’s page signals that at least one credible institution does not expect the vendor to vanish.
6. Oracle’s rivalries were personal—and sometimes ended in acquisition
Marc Benioff may have recognized Ellison’s recurring rupture with protégés and departed before the next vesting date forced a choice between permanent enmity and friendship. His relatively cordial exit was unusual; Salesforce then became part of technology’s next generation.
Tom Siebel represented the opposite extreme. After Siebel accused Ellison of being willing to kill customers’ dogs to make a sale, he appeared with his dog wearing protective armor. Andrew recalled Ellison saying that, with one bullet, “it wouldn’t be your dog” he was worried about. Oracle ultimately bought Siebel for roughly $6 billion—a victory lap over a bitter rival.
The PeopleSoft hostile bid supplied the closing irony: Ellison had spent the book attacking Microsoft over its antitrust case, then Oracle entered its own landmark DOJ case. Once again, the behavior he denounced elsewhere appeared inside his own strategy.
7. Oracle produced leaders without the usual “mafia” catalyst
Hobart’s business-mafia model starts with premature exit. A cohort of roughly 200–250 people has money, unfinished ambition, and a social network saturated by colleagues; PayPal reinforced this by asking recruits to name their five smartest friends.
Tiger Management fit the pattern after bad macro bets at peak AUM fed into a wind-down. Its people could fund new attempts, did not immediately need jobs to pay rent, and still had something to prove.
Decentralization provides another route. GE division heads owned P&Ls and performed miniature CEO jobs, while Palantir’s forward-deployed engineers balance technical constraints, customer politics, opex, and capex as if operating temporary companies inside clients.
Google’s gradual departures created less network density, while Apple and Tesla offered few obvious parallels to Oracle’s executive diaspora. Andrew’s provocative explanation for Oracle’s exception was Ellison himself: by repeatedly “exiting everyone early,” he may have manufactured the pressure a normal long-lived company lacks.
8. Safra Catz won by fitting Oracle’s next problem set
The book’s hindsight failure is stark: its author portrays Catz as intensely loyal but lacking what Oracle’s next CEO would require. Ellison then lists two internal possibilities and an ideal profile sharing little with her eventual role.
Catz nevertheless became the communications gateway around Ellison. Hobart saw possible political leverage: because communication ran through her, she could know where performance was weakening and “where the bodies are buried.” His baseline interpretation was simpler: she was an exceptionally organized operator.
Hobart’s baseline was that Oracle at scale needed fewer “big bold dreamers” and more discipline around buybacks and growth investment. Her later pairing with Mark Hurd, followed by her becoming sole CEO after his 2019 death, also meant she did not initially inherit the role alone.
Hobart’s Apple analogy generalized the succession logic: replacing Steve Jobs sounded like a design question, yet Apple’s next needs favored operational improvement, lower working capital, and outsourced production. “You probably always want to have successive differently good CEOs,” because the unsolved problems reflect the prior leader’s weaknesses.
9. Ellison’s possible outlier skill was translating code into institutions
Structurally, Soft War is not a cradle-to-career biography. Its plot spans roughly six quarters after the dot-com bust, then uses flashbacks and Ellison’s argumentative footnotes; the author rides in his cars and planes, giving it the feel of “Gonzo journalism.”
Ellison’s adoption, difficult relationship with his adoptive father, and possible chip on his shoulder arrive surprisingly late, after vast stretches of boat racing. Andrew contrasted that with Buffett’s childhood, where extraordinary commercial ability was obvious early; Ellison felt more mortal, charismatic, and shaped by timing and opportunity.
What changed Hobart’s view was Ellison’s genuine vision. He had previously assumed Oracle’s imagined future was simply “the quarter they’re guiding for,” yet Ellison anticipated centralized data centers, networked computers, browser-mediated use, and software standardization.
His most arrogant instruction proved the most prescient: do not customize software around the business; change the business around the software. Like buttons, status updates, and platforms that become verbs now shape thought and behavior—users remain active, but they are “riding on tracks that are written in code.” Hobart’s candidate for Ellison’s rare skill was translating between technical constraints and human organizations, then possessing the will to enforce the result.