What Really Happened to Theranos: The Fraud Mechanics, Not the Drama

The Theranos fraud was, at its core, a product that did not exist being sold as a product that did. Between 2003 and 2018, Elizabeth Holmes and her partner Ramesh "Sunny" Balwani raised over $700 million from investors, built a company to a peak valuation of $9 billion, entered into major commercial partnerships with Walgreens and Safeway, and ran blood tests on real patients through American pharmacies — all while knowing that the central technology they were selling was broken. Patients received results from tests run on machines the company did not disclose, diluted with techniques that magnified inaccuracies, filtered through a quality control system designed to hide failures rather than flag them.

This article does not cover the personality of Elizabeth Holmes, the board of directors she assembled, or the cultural dynamics of Silicon Valley that made her story possible. Those stories exist elsewhere in abundance. This covers the specific, technical, financial, and institutional mechanics of how the fraud actually operated — because understanding the mechanics is the only way to understand how it ran for fifteen years without being stopped.

Holmes founded Theranos in 2003, aged nineteen, after dropping out of Stanford's School of Engineering. The founding proposition was genuinely compelling: replace the traditional venous blood draw — a needle in the arm, multiple vials, a laboratory far from the collection point — with a finger-prick test that filled a tiny proprietary container called a nanotainer. From that single, small blood sample, the company claimed its device could run more than 200 diagnostic blood tests with the same accuracy as conventional laboratory analysis, returning results within hours, at a fraction of the cost charged by established labs like Quest Diagnostics and Labcorp.

The device that would perform these tests was called the Edison. It was presented as a miniaturised laboratory — a self-contained, proprietary machine small enough to sit on a countertop that would process the nanotainer sample and produce clinical-grade results for everything from a basic cholesterol panel to HIV screening, thyroid function, cancer markers, and blood coagulation tests. The pitch was that this technology would democratise healthcare — making accurate diagnostics accessible, affordable, and painless for everyone.

It was, in its entirety, a fabrication.

The Edison device was never capable of what Holmes and Balwani claimed. This was not a case of a promising technology that fell short of its potential during development. It was a machine that failed basic laboratory standards and never approached clinical reliability for the tests it was supposed to run.

Theranos's own whistleblowers, regulatory inspectors, and ultimately its own lab directors confirmed the same picture: the Edison consistently failed quality control tests. The machines produced results that varied wildly between tests on the same sample — what laboratory scientists call high coefficient of variation — which is the technical definition of an unreliable instrument. In coefficient of variation studies, deviant results were the norm rather than the exception. Former lab associate Erika Cheung, testifying at Holmes's criminal trial, summarised the failure rate with unusual clarity:

Of the more than 200 tests Theranos advertised, the Edison could reliably perform approximately 12. Of those, only one — the herpes test — received FDA approval. Every other test on the company's patient-facing menu required a different approach entirely. And that different approach was the core of the fraud.

Because the Edison could not run the tests on the company's menu, Theranos needed a different solution. The solution they chose was to run most tests on conventional laboratory machines purchased from established manufacturers — primarily Siemens — while telling patients, physicians, investors, and commercial partners that the results came from Theranos's proprietary technology. They were marketing the importance of their Edison machines to the public while actually running samples on third-party machines.

This substitution was the central operational deception. When a patient in a Walgreens Wellness Center had blood collected and submitted to Theranos for testing, the test report they received listed the company's Newark, California laboratory as the testing site and described the process in terms consistent with Theranos's proprietary technology. In reality, for the vast majority of tests, a Siemens analyzer or another commercial device was doing the work — the same type of machine that Quest Diagnostics or any other conventional laboratory would use.

Theranos also had a specific logic it deployed internally to justify the substitution during regulatory processes. When running proficiency testing for clinical laboratory accreditation — the external verification process that all licensed labs must pass — Theranos ran the proficiency samples on the third-party analyzers rather than the Edison machines. Their stated rationale was that the Edison's technology was so unique that no peer group existed for comparison. The practical effect was that their accreditation was based on machines they were not publicising, and the machines they were publicising were never independently validated for accuracy.

The nanotainer — the small finger-prick collection device — produced a sample volume far smaller than standard laboratory protocols require. Conventional venous blood draws yield several millilitres of blood per test. The nanotainer yielded a fraction of that. This created a fundamental problem: the Siemens machines Theranos was secretly using were designed and calibrated for standard sample volumes. Running a sub-standard volume through them would produce unreliable results.

Theranos's engineering solution was to dilute the blood samples with a saline-like solution to bring them up to the volume the machines needed. This introduced a second layer of inaccuracy that compounded the first. Dilution lowers the concentration of the substances being measured — the analytes — below the sensitivity thresholds the machines were qualified to detect. Tiny measurement inaccuracies, which are inherent in any instrument at the margins of its detection range, get magnified when the sample is diluted. A small error in measuring a normal-concentration sample becomes a larger error when the concentration has been artificially reduced.

The consequence was that even when the Siemens machines were functioning correctly, the diluted samples produced results that diverged from what a standard test on a properly drawn venous blood sample would show. Patients received test values that reflected not their actual blood composition but their actual blood composition altered by dilution and the amplification of measurement error. The margin of inaccuracy was not trivial. Former employees confirmed that Theranos was aware of this problem and proceeded regardless.

Even with the Siemens substitution and the dilution methodology, quality control remained a persistent problem. Quality control in a clinical laboratory involves running known reference samples — samples with a certified, established value — through the machine and checking that the results fall within an acceptable range. If the results fall outside that range, the test run is flagged as failed, the results are held, and the cause of the failure is investigated before patient results are released.

At Theranos, quality control tests routinely failed. The company's response was not to investigate the failures or to withhold results. It was to build a system to remove the failing data points. Erika Cheung testified that the company used an "outlier deletion" system to cherry-pick the best data points from quality control runs and use only those to demonstrate compliance. Data points that fell outside acceptable ranges were flagged as outliers and deleted from the quality control record.

Theranos periodically invited high-value visitors — investors, potential partners, board members, government officials — to witness demonstrations of the Edison device. These demonstrations were central to the fundraising operation. Watching the machine run a test and produce a result was the most direct validation an investor could seek short of independent laboratory testing, which Theranos consistently resisted and refused.

Court testimony from Daniel Edlin, a former Theranos project manager, revealed what actually happened during these demonstrations. The Edison machines used for investor visits ran a demo application that was programmed to hide any failure messages the machine generated. When the machine produced an error — which, given the Edison's reliability record, occurred regularly — the demo app suppressed the error code and displayed a normal-looking result screen. Additionally, the machines included what was described in testimony as a null protocol: a mode in which the device could be activated and appear to be running a test while performing no actual sample analysis at all.

The machines were, in the most literal sense, props. A visiting investor who watched an Edison run a blood test and produce a result was watching a performance. The device's actual analytical capability was irrelevant to what they saw. The result on the screen was not derived from the sample. It was a display state.

The technical deception enabled the financial deception, but the financial deception required a separate set of fabrications that went beyond the laboratory. To attract and retain investors at the valuations Holmes was targeting, Theranos needed to demonstrate not just working technology but a growing business with real commercial revenues.

The numbers Holmes presented to investors were fabricated at a scale that, when later revealed, was described by the SEC as an "elaborate, years-long fraud." The most stark single data point: Theranos claimed revenues of approximately $100 million for 2014. The company's actual revenues for that year were approximately $100,000. The stated figure was one thousand times larger than the real one. Holmes was not rounding aggressively or projecting optimistically. She was inventing a number and presenting it as a fact to investors committing tens of millions of dollars on the basis of it.

The investor fraud required not just fabricated revenue numbers but fabricated credibility. Holmes understood that institutional validators — major pharmaceutical companies, military customers, established healthcare partners — were the single most powerful signal of legitimacy for a healthcare technology company. No sophisticated investor would stake tens of millions on an unproven device without some evidence that the technology had been examined and endorsed by entities with the expertise to evaluate it.

Theranos's approach to this problem was to take real but limited commercial relationships and represent them to investors as something categorically different. Pfizer and GlaxoSmithKline had engaged Theranos for limited clinical trial work — early-stage, exploratory relationships that in no way constituted validation of the Edison's diagnostic accuracy. Holmes put their logos on investor documents as if they had. The Cleveland Clinic and Capital BlueCross entered partnerships that were presented to investors as endorsements by respected healthcare institutions. The DoD pilot programme — which had found the technology unsatisfactory — was presented as an active operational deployment.

This pattern exploited a structural vulnerability in how sophisticated investors evaluate early-stage technology companies. If a firm as rigorous as Pfizer appears to have endorsed the technology, most investors will reduce the intensity of their own due diligence on the assumption that Pfizer's scientists have done it for them. Holmes effectively outsourced the credibility of the investment case to institutions she had neither the permission nor the basis to invoke.

For a fraud of this scale and duration to persist, internal dissent had to be suppressed. Theranos had many highly qualified scientists and laboratory professionals who understood, at various points, that the technology was not performing as represented. Some of them raised concerns. The company's response was systematic and consistent.

Employees operated under exceptionally restrictive non-disclosure agreements that made it legally dangerous to discuss the company's internal operations, technology, or practices with anyone outside the organisation — including, in some interpretations, regulatory authorities. Staff who raised concerns were, as characterised by their own testimony, "usually marginalized or fired." Holmes and Balwani ran what former employees described as a "dysfunctional corporate culture" of "secrecy and fear." The company's internal culture treated information security as a near-absolute operational priority, which served the dual purpose of protecting genuine intellectual property and suppressing awareness of the technical failures.

The most dangerous internal dissenters were the company's own laboratory directors — medical professionals with legal obligations to ensure the accuracy of tests being reported to patients. Adam Rosendorff, a medical doctor who served as lab director, left the company in 2014 and later became a key prosecution witness. His successor, Sunil Dhawan, testified that Holmes blocked him from implementing changes needed to address quality concerns. A later lab director, Mark Pandori, also left after conflicts over the validity of the results being reported. Each departure removed a potential internal check. The pattern was consistent: laboratory professionals who refused to validate what the company was doing were cycled out.

The fraud unravelled through three converging forces that arrived, in sequence, between 2015 and 2016: a federal regulatory inspection, a journalist, and two former employees who chose to speak.

In October 2015, Wall Street Journal reporter John Carreyrou published the first of a series of investigations into Theranos, based on information provided by sources including Tyler Shultz — a Theranos employee and grandson of former Secretary of State George Shultz, who was on the company's board — and Erika Cheung. Carreyrou's reporting established the core facts that later formed the foundation of the government's case: the Edison's limited capabilities, the reliance on third-party machines, the accuracy problems, and the gap between what the company was telling the public and what its own scientists knew internally.

Separately, the Centers for Medicare and Medicaid Services (CMS) — the federal agency responsible for regulating clinical laboratories — conducted an inspection of Theranos's Newark laboratory and issued findings of serious deficiencies. The CMS inspection found violations of federal laboratory standards across multiple categories. The agency determined that Theranos's laboratory posed "immediate jeopardy to patient health and safety" — the most serious finding CMS can issue. In response, Theranos was required to void all results from its Edison devices for 2014 and 2015 and to issue corrected reports to the doctors and patients who had received those results.

Most analysis of the Theranos case focuses on the financial victims — the investors who lost approximately $945 million in aggregate, or the institutional partners who wrote off hundreds of millions in partnership investments. Rupert Murdoch alone lost $125 million. The Walton family lost a comparable sum. These are the figures that attract the most attention because they are the figures that drove the criminal prosecution.

The patient harm is harder to quantify but no less real. Patients were misdiagnosed with conditions ranging from diabetes to cancer, based on Theranos results. Others were given clean results for conditions they actually had. Former lab director Sunil Das testified about a specific example: a Theranos test for prostate cancer was producing positive results in a large number of women — a biological impossibility that was a direct indicator of instrument error. When Das raised this with Holmes as evidence of the Edison's "error propensity," she suggested, based on a study she had read, that the women might have a rare form of breast cancer. Das testified that he "did not think that explanation was plausible."

The legal framework of the criminal prosecution did not ultimately produce convictions on the patient fraud counts. Holmes was acquitted on the charges relating to patients at trial — the jury found insufficient evidence that she had intentionally defrauded individual patients. But acquittal on those counts is not a finding that the patient harm did not occur. It is a finding that the criminal intent standard required for conviction was not met on those specific charges. The 945,000 patients who received unreliable results, the physicians who treated them on the basis of those results, and the clinical decisions made on that data are an established matter of record.