Insight, Industrial Policy
A Nigerian programme is building fabrication and testing capacity for hardware founders. Indonesia is planning 24 GW of solar and wind plus 10 GW of storage by 2034. Both are constrained by the same thing, and it is not money. It is the time it takes to learn.
Two announcements in today's briefing appear to belong to different conversations.
Off-grid energy investor All On has launched a twelve-month Innovation Ecosystem Development Programme, backed by the Global Energy Alliance for People and Planet. Ten Nigerian clean-energy ventures will receive technical expertise, prototyping facilities, fabrication equipment, venture-building support, and mentorship, rather than another pitch competition and a cash prize.
Separately, a Tara Climate Foundation assessment reports its Asian focus markets adding on the order of 35 GW of renewables in 2025, with solar capacity up in the mid-teens and wind in the low-teens as a percentage. Indonesia's latest electricity plan (RUPTL) targets roughly 24 GW of solar and wind plus 10 GW of storage by 2034, while ASEAN countries align sustainable-finance taxonomies and governments refine auctions, power-purchase agreements, and corporate procurement rules.
One is a small programme for ten companies. The other is a regional capacity build measured in tens of gigawatts. They are the same problem at two scales, and the reason to put them together is that neither is short of money.
Sanjaya Lall's 1992 work on technological capabilities and industrialisation set out something that development practice keeps having to relearn.
Technology does not arrive as a working capability. A firm that buys equipment has bought equipment. Turning that into the ability to design, produce, adapt, improve, and troubleshoot is a separate process, and it is expensive, incremental, largely tacit, and accumulated through repetition and failure over years. Lall's framework locates capability in the interaction of three things: incentives that make learning worth doing, human capital able to do the learning, and institutions that support it. Remove any one and the other two do not compensate.
The corollary is the part that gets skipped. Capability cannot be transferred by giving somebody the equipment, the training, or the money. It can only be built by doing the work, repeatedly, with tolerance for the failures that constitute the learning.
This applies with as much force to states as to firms. Designing a renewable auction that clears at a sensible price without attracting unqualified bidders is a skill. Drafting a power-purchase agreement that a lender will accept and a regulator will defend is a skill. Managing an interconnection queue, running a permitting process that is fast and not corrupt, negotiating a benefit-sharing agreement with a community that has been promised things before: all skills, all acquired by doing them badly first.
Set the All On programme against that.
The diagnosis behind it is exactly right, and better than the one usually offered. African climate-technology discussions default to a shortage of venture capital. The earlier bottleneck is engineering infrastructure. A hardware founder needs a laboratory, testing equipment, fabrication capacity, and access to engineers before there is anything an investor could plausibly finance. No amount of capital fixes the absence of a machine shop. This is the domestic version of the argument we made about the electrical-equipment supply chain: the assembly, testing, and commissioning layer is where capability can realistically be built, and it is scarce everywhere. Identifying that and building the layer is a considerably more useful intervention than another accelerator.
The tension is between the diagnosis and the timeframe. Firm-level capability formation, on the evidence Lall assembled and everything since, runs on a scale of years to a decade. Twelve months is enough to produce prototypes and some trained people. It is not enough to produce firms with production capability, and nobody should expect it to be.
That is not a criticism of the programme, which has to have a duration and twelve months is a reasonable first one. It is an argument about what success should be measured against, and it leads somewhere counterintuitive.
Ten ventures will be selected and most will fail, which is the normal and correct outcome of early-stage selection. What should persist is the physical capacity: the prototyping equipment, the test rigs, the fabrication capability, and the engineers who know how to use them. If the facility is still operating in year five and has served five cohorts, the programme worked. If the equipment is idle after the grant ends, it did not, regardless of how the ten companies performed.
Failed ventures that leave behind trained engineers are a positive return. An engineer who spends a year designing a solar controller, watches the company fail, and takes that experience to the next firm has added to national capability. On a venture-return metric this is a write-off. On a capability metric it is precisely the mechanism by which industrial capability accumulates, and it is how Taiwan and Korea built theirs.
This matters because capability programmes are increasingly funded through venture logic, with cohort outcomes, portfolio returns, and graduation rates as the reported metrics. Those metrics actively penalise the thing the programme exists to produce. A programme judged on how many of its ten companies raised a Series A will optimise for selecting companies that were nearly investable already, which is the opposite of building the layer that was missing.
24 GW of solar and wind plus 10 GW of storage by 2034 is a target. Targets are the fastest thing in the transition to produce and the least informative.
What determines whether that number is met is the delivery architecture underneath it, and the Tara assessment's attention to that layer is the most useful thing in the item. Auction design. Power-purchase agreement terms that are bankable without being extortionate. Grid codes for storage, which most systems in the region do not yet have. Interconnection-queue management, which becomes the binding constraint the moment auctions start clearing. Corporate procurement rules. Taxonomy alignment across ASEAN so that a project qualifying as sustainable in one jurisdiction is not disqualified in another.
Every item on that list is a state capability, and every one is built by drafting a rule, watching it fail in practice, and drafting a better one. There is no shortcut and there is no consultant who can supply it, because the knowledge that matters is knowledge of how the rule interacts with this grid, these developers, this legal system, and these communities.
Storage in particular is going to expose this. 10 GW of storage in a system whose market rules were written for dispatchable thermal generation and intermittent solar is not a procurement exercise. Storage does not fit cleanly into the categories most regulatory frameworks in the region have available. Whether it is generation, load, or a network asset determines who may own it, how it is remunerated, and whether it can stack services. Getting that wrong does not slow deployment. It makes the assets uneconomic after they are built, which is worse.
One detail in the Tara assessment deserves to be pulled out and stated on its own, because it is the most consequential thing in today's briefing and it will get the least attention.
The report highlights community participation and benefit-sharing mechanisms around offshore wind and nickel regions, rather than treating social acceptance as something to address after deployment.
That sequencing is the whole argument. Social acceptance addressed after deployment is a renegotiation conducted from a position of weakness, with sunk capital on the table, a construction schedule under pressure, and a community that has correctly worked out it now has leverage it did not have before. It is the most expensive possible moment to have the conversation, and it produces agreements that satisfy nobody and hold badly.
Social acceptance addressed before deployment is a design input. It changes siting, employment terms, revenue sharing, land arrangements, and grievance procedures while changing them is still cheap. It also produces information: a community that has been consulted properly will tell you things about water, land tenure, seasonal labour, and previous broken promises that no feasibility study will surface. That is what field research is for.
Indonesia's nickel regions are the sharpest version of this. The energy transition and the industrial policy are inseparable there, the social terms of extraction are being set now, and they will be very difficult to revise later. Our fieldwork in Lombok works at exactly this level, on what a technology deployment means to the people who live where it lands, and it is consistently the layer that project documentation describes least well.
This is also the capability that takes longest to build and is quickest to destroy. A state or a developer can acquire technical competence in a decade. Trust with a community that has been promised things before and seen them not arrive is built over the same period and can be lost in a single badly handled season.
If capability is the slow part, then sequencing is the only real lever anybody has, and the implication is uncomfortable for how these things are funded.
Start the slow things first, even where the fast things are more visible. The fabrication facility, the regulatory drafting capacity, and the community relationships all take years and none of them produce an announcement. They should nonetheless begin before the capacity target, not after it, because the target's feasibility depends on them.
Fund capability on capability timescales. A twelve-month programme building a decade-long asset needs a stated path to years two through five at the point it launches, or it is a pilot that will be reported as a success and then stop. This is a funder decision rather than a programme decision.
Judge capability programmes on the capability, not the cohort. The instruments most likely to detect this are the open-ended ones, for the reason set out in The Benefits Nobody Was Looking For: a results framework written around ten companies cannot see an engineer who left the eleventh. Is the facility still running? How many engineers have passed through it? What can be fabricated locally now that could not be before? How many failed ventures released trained people into other firms? These are the returns. Company survival rates are a poor proxy and they push the programme in the wrong direction.
Treat community trust as infrastructure with a maintenance budget. It depreciates, it is expensive to rebuild, and almost no project budget contains a line for maintaining it after commissioning. See TRL and SRL: The Two Axes of Readiness for the framework the Lab uses to keep the social side visible in a technical assessment.
The Lab works on this in local manufacturing and energy access. The argument that the social layer has to be understood before the technical one is not a methodological preference. It is a claim about which constraint binds first, and everything above is the reason we hold it.
If you are designing a capability programme and want the right things measured from the start rather than the convenient ones, tell us what you need to know.
This is an independent insight piece by Transitions Lab. For the Lab's applied work, see Local Manufacturing & Supply Chains and Energy Access & Off-Grid Systems. See also Paying for Power You Curtail, which reads the same slow-implementation problem at the level of a single national grid, The Value Is in the Second Step on why operations capability is the part that has to be built rather than procured, Behind the Border on why the behind-the-border capabilities decide whether regional value chains form, The Ban Is Not the Policy on the decade of local integration that determines whether a plant is capacity or capability, Symbiosis Does Not Arrive on a Site Plan on the institutional layer that a shared industrial system has to accrete, The Municipality Is the Instrument on the same capability question inside a small council, The Mandate Is the Mine on what a licensed synthetic fuel plant retains in country, The Survey Is the First Act of the Mine on why interpretive capability decides whether a state databank is an instrument or a filing cabinet, and Strategic Is Not the Same as Financeable on the distance between a policy commitment and an operating industry. To discuss a study, see Contact.