CSIR was established in 1945 to produce innovation South Africa cannot buy. It learned instead to produce consulting South Africa will pay for. These are different activities, and only one of them compounds.
CSIR does not fail to be innovative. It succeeds at being a consultancy. These are mutually exclusive outcomes when an institution was built to do the second thing while producing the first.
The commercial incentive structure selects, rationally and systematically, against the kind of work that produces genuine innovation. Short timelines, specified outcomes, paying clients, low risk. These are not failures of culture. They are the logical outputs of a funding model pointed the wrong way.
CSIR is a Schedule 3B public entity: largely self-funding, dependent on competitive contracts from government and private clients. The contract structure is the innovation killer. Deliverable-based contracts systematically reward low-risk, short-horizon, specified-outcome work.
CSIR trains researchers at public expense, then pays them below the market rate that its commercial clients — the same organisations buying CSIR consulting — will offer them the moment they are trained.
In 2016 the CSIR CEO resigned publicly, citing political pressure to favour Huawei in a R116 million supercomputer procurement. The pressure allegedly came through the Minister of Science and Technology and the ANC treasurer-general. The CEO named the pressure explicitly. He left. The system continued.
CSIR patented the appetite-suppressing properties of the Hoodia plant, knowledge held by the San people for centuries, without recognising their traditional claims. The patent was sold to Unilever. A class action eventually produced a revenue-sharing agreement. The original act of appropriation was institutional reflex, not aberration.
CSIR has defence, security, and advanced manufacturing divisions. The sanctions era produced civilian spillovers from defence research in most comparable economies. CSIR’s defence work has produced no equivalent. The institutional separation between CSIR defence research and the South African tech economy is total.
Real innovation is non-linear and time-indeterminate. Contract-based R&D funding is linear and time-bounded. The mismatch is not manageable at the margin; it is structural. An institution that must justify every expenditure against a deliverable cannot take the risks that innovation requires.
CSIR produces intellectual property. In theory, this IP should seed South African companies. In practice, the commercialisation pathway — from research output to tradeable product to scaled company — is underdeveloped, underfunded, and structurally disconnected from the South African capital market.
CSIR turned 80 in 2025. The institution has never satisfactorily answered: what is the compound impact of eight decades of public investment on South Africa’s industrial and technological base? Not what it has produced. What has compounded.
CSIR’s dysfunction is not a management failure. It is a funding architecture failure. Four structural mechanisms hold the commercial-capture equilibrium in place.
Every deliverable-based contract selects for a specific type of research: short-horizon, low-variance, client-specified. Over 80 years, this selection pressure shapes not just the portfolio but the researcher base. CSIR hires for, develops, and retains the profile that wins contracts. That profile is not the profile that produces breakthrough innovation.
The parliamentary grant that CSIR receives is insufficient to fund the long-horizon, high-risk work that would make commercial contract dependency unnecessary — but large enough to sustain the governance structure, infrastructure, and overhead of an institution that still presents itself as a national R&D body. The grant funds the frame. The contracts fill the picture.
South African R&D salaries in the public sector are constrained by public service remuneration frameworks. The private sector, particularly international firms and multinationals operating in South Africa, pays a significant premium for exactly the profile that CSIR trains. The institution spends three to seven years developing a researcher and then watches them leave for a multiple of their CSIR salary.
The 2016 Sibusiso Sibisi resignation, naming political pressure to favour Huawei in a major procurement decision, is the most visible event in a pattern that is structural rather than episodic. CSIR’s CEO is a political appointment in effect, if not in formal procedure. Political procurement pressure on a research institution does not only distort individual contracts; it distorts the research agenda over time.
The commercial mandate did not corrupt CSIR. It completed it — into something its founders would not recognise.
The capture-by-optimisation argumentThree events that illustrate the structural diagnosis. Each one is individually explicable. Together they describe an institution whose governance has been repeatedly compromised at exactly the moments when independence mattered most.
CSIR CEO Sibusiso Sibisi resigned and published an open letter naming political pressure from Science and Technology Minister Naledi Pandor and ANC treasurer-general Zweli Mkhize to favour Huawei Technologies in the procurement of a R116 million supercomputer for CSIR’s High Performance Computing Centre. The pressure, Sibisi wrote, included irregular conduct that raised serious concerns about procurement integrity at a national science institution.
The significance is not the specific procurement. It is that a sitting CEO of a public science institution felt compelled to resign publicly and name the interference explicitly rather than manage it internally. The inference is that internal management had already failed to contain the pressure. The supercomputer was eventually procured. The governance architecture that allowed the pressure to be applied was not reformed.
CSIR researchers identified the appetite-suppressing properties of Hoodia gordonii and filed patents without recognising the San people’s traditional knowledge claims, despite centuries of documented San use of the plant as a hunger suppressant during long hunts. CSIR subsequently licensed the patent to Unilever, which marketed Hoodia products internationally. The South African San Council filed a formal complaint in 2001. A revenue-sharing agreement was eventually reached in 2003, with the San receiving 6–8% of royalties.
The institutional lesson that was not drawn: the Hoodia case was treated as a knowledge management and intellectual property failure. It was not treated as evidence that the commercial mandate, applied without ethical constraint, produces extraction rather than innovation. The same incentive structure that produced Hoodia continues to operate. The mechanism has not changed.
CSIR employed approximately 2,298 staff as of its most recent reporting, of whom 1,617 were scientists and engineers. That figure is not the problem. The problem is the ratio of staff who trained or developed skills at CSIR and subsequently left for the private sector, international institutions, or emigration. CSIR does not publish this figure. No public accountability mechanism requires it to. The institution is permitted to describe its human capital investment as a national contribution while bearing none of the accounting cost of the return on that investment flowing elsewhere.
A realistic estimate of CSIR’s contribution to South Africa’s innovation economy would need to include the value of skills that compounded internationally rather than domestically. That estimate has never been attempted, which is itself a finding.
Read the mandate language and the gap between what CSIR presents itself as and what its incentive structure produces becomes impossible to ignore.
The word that appears most in CSIR’s self-description is “directed.” Directed research. Directed multidisciplinary investigation. Research directed at national priorities. The choice of that word is diagnostic. Directed research is research in which someone else specifies the direction. That is precisely the commercial-contract model the institution has been captured by. The mandate does not describe innovation. It describes responsive service delivery for whoever holds the contract.
The 1945 Act used a different vocabulary. “Foster.” “Benefit of the Republic.” These are generative terms, not transactional ones. They describe an institution that grows something, not one that delivers something. The strategic rewriting of that language — accumulated across successive planning cycles into the current “directed research aligned with client priorities” formulation — is the intellectual audit of 80 years of commercial capture. The vocabulary is how the institution stopped being a different kind of thing.
Each of CSIR’s stated values, set against the commercial-capture diagnosis, reveals a specific dissonance.
Excellence in the scientific sense means doing work that would not be done without you, that advances the frontier rather than applies it. Excellence in the commercial-contract sense means delivering what the client specified, on time and within budget. The second definition does not produce Nobel laureates. It does not produce the foundational work on which industries are built. It produces CSIR research reports that government departments reference in procurement decisions.
South Africa has no equivalent of the fundamental research breakthroughs — the transistor from Bell Labs, the internet from DARPA-funded universities, the human genome from a network of publicly-funded institutions — that have seeded entire economic sectors. CSIR has been here for 80 years. The absence of an equivalent contribution is not explained by the talent. It is explained by the incentive structure that has consistently discouraged the kind of work that produces it.
Tangibility and measurability are contract-language values. They describe the kind of outputs that justify a fee note. They do not describe the kind of outputs that create new industries. The most valuable research outputs in history — Maxwell’s equations, Fleming’s penicillin observation, Shannon’s information theory — were not measurably tangible at the moment of production. Their impact was only legible decades later, once the applications had propagated through the economy.
CSIR’s impact reporting framework counts patents, publications, contracts won, and clients served. It does not count companies seeded, industries created, or technologies that scaled into the economy without CSIR’s name on the downstream application. That is not a reporting gap. It is a theory of impact that has been written to match the evidence that the current model produces, rather than the evidence that a genuinely innovative institution should produce.
The collaboration model CSIR describes is the contract model renamed. Co-creation with a paying client is consulting. The genuine collaboration that compounds — the cross-institutional, long-horizon, non-specified-outcome collaboration that builds bodies of knowledge — requires partners who are not paying clients and timelines that no contract can encompass.
CSIR’s most important collaborators, on the current model, are its contract clients. Its most important potential collaborators — the South African university system, the diaspora of trained researchers who left, the African continental institutions working on adjacent problems — are the least resourced actors in the science system. The collaboration value describes relationships of commercial dependency, not relationships of intellectual compounding.
CSIR cannot reform itself within the current funding architecture. The reforms that matter are not management reforms. They are mandate and funding architecture reforms.
The parliamentary allocation should be explicitly ring-fenced for research that cannot be funded by commercial contracts: high-risk, long-horizon, non-specified-outcome work. A minimum of 40% of parliamentary funding should be untied to any client or deliverable requirement. This is not generosity. It is the condition for genuine innovation capacity.
Every CSIR research programme above a defined IP threshold must produce a commercialisation assessment: is this licensable, spinout-worthy, or patent-ready? The assessment is not a bureaucratic exercise; it is connected to a ring-fenced commercialisation fund with the authority to invest equity in spinouts. CSIR creates companies, not just reports.
A specific, Treasury-approved salary band for senior CSIR researchers that is benchmarked against the international market rate, not the South African public service salary framework. Losing a trained researcher to a private consultancy is a negative return on the public investment in their development. The cost of the retention premium is less than the cost of the training cycle.
CSIR procurement decisions above R10 million should require sign-off from an independent procurement committee with no ministerial representation. This is not anti-political; it is science-system hygiene. A research institution whose major procurement decisions are subject to political direction is not a research institution. It is a government department with laboratory equipment.
CSIR should be required to produce a five-year compounding impact assessment that tracks not only what it has produced but what has scaled from those outputs into the economy: companies formed, technologies adopted at scale, researchers whose work built on CSIR foundations. An institution that cannot account for its compounding impact is not measuring the right things.
The post-Hoodia framework for indigenous knowledge recognition has been revised but not institutionalised. A standing Indigenous Knowledge Review Committee, with San and other community representatives and independent legal authority, must review any CSIR patent application that draws on traditional knowledge. The Hoodia settlement was a negotiated outcome after the damage was done. The protocol must operate before the patent is filed.