If You Don't Build It, They Can't Come
Build the enabling layer before latent demand can become a market
- Difficulty
- Expert
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 99%
This model reverses the conventional requirement to prove a market before building a product. First identify a technically possible future whose participants are blocked by missing infrastructure. Build the enabling platform, make it usable enough for researchers and developers to experiment, and maintain the investment while watching for independent breakthroughs. The mechanism is not blind faith that customers will arrive; it is recognition that customers and applications cannot exist until somebody creates the prerequisite layer. NVIDIA followed this logic with CUDA years before deep learning supplied a mass-market workload. The approach remains highly risky because timing and demand are uncertain, so it requires financial endurance, continuous external sensing, and the ability to mobilize quickly when a validating use case appears.
Origin
Extracted from Acquired's account of Jensen Huang funding CUDA before a commercially sufficient market for general-purpose GPU computing existed.
Core principles
- 01Some markets cannot emerge until enabling infrastructure exists
- 02Absence of present demand does not prove absence of future demand
- 03Foundational platforms require patience before applications appear
- 04External breakthroughs can unlock deliberately created capabilities
How to run it
- 1
Define the blocked future
Describe the valuable application or ecosystem that cannot emerge with current infrastructure. Identify the specific technical constraint preventing participants from acting.
Pro tip Look for expert users already forcing unsuitable tools to perform the job.
Watch out Do not confuse a merely unpopular product with a genuinely missing enabling capability.
- 2
Build the prerequisite layer
Create the minimum complete platform that makes the previously impossible activity feasible. Include enough tooling and abstraction for outsiders to experiment without mastering the hardware.
Pro tip Invest in developer tools, documentation, libraries, and support alongside the core technology.
Watch out A low-level component alone may be technically impressive but commercially unusable.
- 3
Fund the waiting period
Set explicit limits for investment, runway, milestones, and strategic patience while the market develops. Continue strengthening the platform without assuming immediate revenue.
Pro tip Use existing profitable businesses to finance the option when possible.
Watch out An unlimited commitment without milestones turns strategic patience into uncontrolled spending.
- 4
Watch for independent validation
Track researchers, developers, and adjacent industries for applications that suddenly become practical on the platform. Treat unexpected usage as evidence rather than noise.
Pro tip Maintain direct relationships with technically advanced early adopters.
Watch out Do not become so attached to the original target market that you miss a larger emergent one.
- 5
Concentrate on the breakthrough
When a credible use case demonstrates exceptional performance and broad applicability, align the organization around it. Build specialized libraries and distribution that make replication easy.
Pro tip Convert expert demonstrations into high-level tools for mainstream practitioners.
Watch out A single experiment is not enough; confirm that the mechanism transfers across teams and applications.
In the wild
NVIDIA invested in CUDA from 2006 even though scientific computing alone could not justify the undertaking. In 2012, AlexNet used CUDA on NVIDIA GPUs to produce a dramatic ImageNet improvement. NVIDIA recognized that deep learning was the high-throughput application its platform had been waiting for and rapidly reinforced it with tools such as CUDNN.
→ A speculative enabling platform became the foundation of NVIDIA's data-center and machine-learning business.
A robotics company believes manufacturers will eventually test factory changes in digital twins. Rather than waiting for a mature market, it builds the simulation runtime, developer tools, and connectors that make such testing possible, while setting funding milestones and recruiting a few advanced design partners.
→ The company creates a defensible option on a market that could not develop without its infrastructure.
Common mistakes
Treating conviction as proof
A founder's belief does not establish that demand will arrive. The bet still needs milestones, survival limits, and independent validation signals.
Building only the technical core
Users cannot create a market around inaccessible infrastructure. Compilers, libraries, documentation, and developer support may be as important as the underlying hardware.
Missing the unexpected application
The eventual market may differ from the original thesis. NVIDIA's decisive opportunity came from deep learning rather than the smaller scientific-computing market initially contemplated.
Is it for you?
Best for
It is best for technically capable companies pursuing platform shifts with potentially large but uncertain downstream markets.
Not ideal for
It is not ideal for companies without the capital, technical depth, or patience to survive a long adoption cycle.
From the transcript
“The phrase he uses is if you don't build it, they can't come.”
“It's like, well, if we don't build it, they can't even possibly come. I don't know if they will come, but they can't come if…”
“Jensen felt like the GPU platform could enable things that the CPU paradigm could not.”
From the episode
Nvidia Part II: The Machine Learning Company (2006-2022)