Max Hodak reveals why deep tech startups fail—not due to bad technology, but poor infrastructure. Learn how procurement, hiring, and speed drive startup succ...
What Kills Deep Tech Startups: Why Infrastructure Determines Speed
Key Insights
- Infrastructure, not technology, kills most deep tech startups. Companies with stellar scientists and engineers fail because they can't manage procurement, hiring, and operational systems efficiently.
- Speed is determined by infrastructure. Faster iteration cycles compound dramatically over time—learning one thing weekly beats a competitor learning monthly indefinitely.
- Procurement and spending control require early investment. Without a structured purchasing system, founders waste time approving individual purchases instead of setting budgets and giving teams autonomy.
- Hiring must follow a consistent, multi-stage process. Distributed voting, phone screens, homework assignments, and interviews prevent bottlenecks and surface better candidates.
- Performance reviews should be continuous and graph-based. Traditional annual reviews miss issues; "Eigenreviews" use ongoing peer feedback weighted by network trust to surface problems faster.
The Real Problem: Why Smart Founders Still Fail
Most people assume deep tech startups die because the technology doesn't work. Max Hodak, CEO of Science (a retinal prosthesis company), has spent 20 years in neuroscience and brain-computer interfaces. His insight: the technology rarely fails. The organization does.
When you're managing hundreds of people and thousands of square feet of physical infrastructure, without systems to track spending, attribute costs, and manage hiring pipelines, strategy disconnects from execution. The company becomes unwieldy.
Procurement: The Unsexy Foundation of Speed
The Problem
Founders often ignore purchasing. You can use a credit card for the first few things. But when your 17th employee needs a $3,000 power supply and asks for approval, you face a trap:
- Deny it → talented employees can't do their jobs while you wait for an auction deal
- Approve it → burn spirals out of control with no visibility into total spending
The Solution: Budget-First Procurement
Instead of approving individual purchases, establish procurement infrastructure:
- Set departmental budgets
- Give employees autonomy within those budgets
- Require vendors to quote and generate purchase orders
- Track all spending through a centralized system
This takes weeks to set up, but it unlocks three critical abilities:
- Employees work faster without bottleneck approvals
- Burn rate stays predictable
- Cost attribution becomes possible (knowing that a wafer iteration costs $40,000)
Without cost attribution, experiments seem free. You can't price products, estimate runway, or make informed decisions about R&D spend.
Hiring: Scaling Beyond Your Network
The Problem
You can hire from your network for the first 10-20 people. But there aren't enough people in your network to staff an entire company. Hiring from the general public without a structured process becomes a time sink.
The Solution: Four-Stage Hiring Process
Science uses:
Initial Voting (~24-48 hours) – An online application feeds to 7-8 employees with similar backgrounds. They vote: "known good," "strong yes," "yes," "no," or "strong no." This distributes screening across the company instead of bottlenecking at the founder.
- Result: Only 17% of applications advance to the next stage
Phone Screen – Assess three attributes: judgment, horsepower (technical competence), and agency (drive to achieve outcomes)
Homework Assignment – Ideally AI-resistant, high-ceiling, scorable with 2-3 metrics. Allows you to spot exceptional candidates even if they use AI tools
Full Interview – By this stage, your conversion rate to offer should be at least 25%, ensuring you don't waste time on candidates who won't be hired
Performance Reviews: Real-Time Feedback Beats Annual Meetings
The Problem
Traditional 360-degree reviews happen once or twice yearly. They surface issues you already knew about but haven't acted on. Timing is terrible—feedback lag is 6-12 months.
The Solution: Eigenreviews
Every 4-6 weeks, employees answer one question via internal software: "Knowing how it turned out—would you hire this person again today?"
The system then:
- Collects all votes across the company
- Weights votes by "voter credibility" (using eigenvector centrality—similar to Google's PageRank)
- Your vote counts more if everyone rates you highly
- Detects voting cliques via dropout sampling (random edge removal iterations)
- Updates continuously with a ~1-month lag
This gives you weekly signals about who's working and who isn't, without traumatic annual reviews.
Speed Separates Success from Failure
Hodak's central thesis: If one team learns one thing per week and a competitor learns one thing per month, the competitor will never matter. Compounding wins decisively.
Speed depends on infrastructure:
- How quickly can your team buy things?
- How efficiently does hiring work?
- How fast do you get performance feedback?
This is not about being smarter. It's about systems. SpaceX, Tesla, and other fast-moving companies invest heavily in internal software—manufacturing platforms, purchasing systems, hiring tools. Commercial ERP systems can't replicate this fit.
The Judgment Problem
As a founder, you cannot delegate judgment. You must make decisions that make sense to you, even when you're alone in that conviction. Early decisions are easy—others advise, you execute. But years in, when hundreds of millions are at stake, there may be no one to ask.
You must:
- Develop sharp judgment in your domain
- Learn continuously from people with excellent judgment
- Accept that "no general principles" exist for startups—each situation is unique
- Recognize that action produces information. When stuck, you must exert action to create new possibilities.
Conclusion
Deep tech startups fail not because the science is wrong, but because founders neglect infrastructure. Procurement systems, hiring processes, and performance feedback mechanisms seem unglamorous, but they determine whether a company scales or collapses under its own weight. Speed is built through infrastructure, and speed determines success. Invest early in systems that let your team move fast, and your odds improve dramatically.
Original source: Max Hodak: What Really Kills Deep Tech Startups?
powered by osmu.app