There has been a lot of talk recently about the “best” place to start a startup. Elad Gil’s controversial thread on Twitter made a compelling case for why the Bay Area should continue to be the location of choice for founders.
I think some theoretical nuance and practical observations are missing from the conversation, which I hope to add here.
I’ll start by saying that if you are a famous second-time founder with an existing network of talented engineers who will immediately leave their jobs to work with you, or you are a genius technical founder who can attract engineers who want exposure to your technical skills, or you have some other ‘unfair’ advantage in recruiting, there is no better place to start a startup than the Bay Area.
Otherwise, there is no particularly good place to start a startup, and life is going to be hard.
Some background on the Bay Area: the Bay Area benefits from what urban economists call economies of agglomeration – that is, benefits that accrue when firms and skilled workers are located in close proximity. This is what Gil was largely talking about in his thread: “[e]ngineers, sales, PMs, others who have worked at breakout companies all based in [the] Bay Area, [e]arly customers (influencers for consumer, buyers for enterprise) all based in tech clusters, [and s]mart angels & venture capital is here, & smart service providers for legal, real estate.” This creates a positive feedback loop, just like network effects in software that all founders are familiar with, causing the magnitude of the benefits to increase as time goes on.
What has ended up happening in the Bay Area is something that, again, should be familiar to anyone involved in the world of venture: a power law distribution of software-related agglomeration effects amongst US cities. In other words, the lion’s share of agglomeration benefits have accrued to the Bay Area, with some comparatively modest amount allocated to cities like New York and Seattle, and a long tail of trivial also-rans. The Bay Area, in other words, has the highest software-relevant talent density of anywhere in the US. It would seem illogical to locate a startup anywhere but the Bay Area, but that only tells part of the story.
The problem is that it isn’t just that benefits have accrued geographically according to a power law distribution – benefits have accrued to firms within that geography according to a power law distribution. In plain terms, a tiny percentage of the startups within the Bay Area are accruing the vast majority of the (sizeable) benefits that are available as a result of the Bay Area’s dominance. This is natural and expected, but potentially problematic for most founders who might choose to locate there: the benefits exist, but accruing them for your own nascent startup is monumentally difficult. I think that mega-successful investors have a hard time seeing this, since they’re disproportionately exposed to the best and most promising startups.
The most difficult part of breaking through the Bay Area benefit-accrual barrier is at the start. Once a startup reaches escape velocity and joins the small percentage of startups who are accruing the vast majority of the benefits, the startup can take advantage of the positive feedback loop and the geography starts to work tremendously in its favor; until then, the geography works against it. Just as in a startup trying to create some sort of network with its product, the hardest part is bootstrapping the beginnings of the network.
The Bay Area looks incredibly attractive when you measure its talent density, but the important metric is the ratio of talent density to exciting startup density. The Bay Area has the highest talent density of anywhere in the US, but also has the highest density of exciting startups to work for. The extremely competitive market makes it extraordinarily hard to attract that talent pool – however dense it is – to your startup. This is true in the later stage as well as the earlier; even billion-dollar Bay Area enterprise startups like Gusto have tremendous difficulty attracting and retaining top-tier engineers, unless headed by ‘celebrity’ founders or working with unusually interesting technology like machine learning at large scale – or both, like Stripe.
What follows is that the ideal place to start a startup is a place with a large, relatively dense talent pool and a comparatively small and unimpressive cohort of early-stage companies. Of course, such a place probably doesn’t exist – otherwise everyone would be starting exciting companies there – though cities can certainly be plotted on a spectrum from best to worst. Most cities have neither a density of talent nor a density of compelling startups (the two are typically correlated), so most should not be considered. And every city at the more desirable end of the spectrum has its own downsides. The Bay Area is expensive and dysfunctional. Seattle is rainy. New York has bad weather and no outdoor life. A meaningful percentage of people would never consider living in LA. Colorado has a winter. Portland is odd, and Austin is in Texas.
There is no good place to start a startup.
But there are ways of arbitraging geography to a founder’s advantage. A founder can (and certainly should) use top-tier Bay Area service providers and raise money from top-tier Bay Area investors in order to get access to their agglomeration effects, and raise capital at, or close to, Bay Area prices – and then locate the startup in an area with as high of a talent to startup ratio as possible, but a lower cost of living. Recruiting will still be incredibly hard, but the startup will have more time to accumulate advantages before running out of money.
Founders with 100% distributed teams should almost certainly live in the Bay Area themselves, if possible – this gives the founder access to the social and knowledge-based agglomeration effects, while insulating the startup itself from the extreme competitiveness for talent in the Bay Area.
And, of course, it isn’t a definitionally bad idea to start a startup in the Bay Area – it might even be the best place. The Bay Area has undeniable advantages in tempo, intensity, and ambition compared to other cities. Likely the biggest advantage at the earliest stage is rapid failure due to high competitiveness and high cost of living; thanks to a low burn rate, founders in non-competitive geographies might waste years on a bad idea or a bad instance of founder-market fit, whereas Bay Area startups run out of cash and talent much more quickly, allowing the founders to move onto a viable idea, or move on with a more traditional career, years before they otherwise would have. Counterexamples certainly exist, too, where startups in the Bay Area implode when they otherwise would have turned into massive successes with a lower burn rate in another region. But the biggest risk of starting a startup in a non-competitive city is the temptation to raise from non-competitive funds, hire non-top-tier talent, keep burn low and languish for years instead of doing something great.
In the end – barring absolutely massive and sudden virality – most startups win when the ability to recruit outsized talent meets unusually strong founder-market fit, and every founder eventually needs an answer to the question of how this can be achieved in their specific case. “Locating our startup in the Bay Area” is not a compelling answer, as the benefits available in the Bay Area are often canceled out by the massively increased competitiveness. The most effective tactics I’ve seen are outsized pay, functional languages (which is an enormous perk for certain engineers), an extremely high bar for talent (which is a bit of a catch-22, but the best perk for engineers is working with other great engineers), and a distinctive, unusually healthy culture.
In any case, recruiting is a problem just like any other – and business is a neverending set of problems, so you’d better like problem solving. The easy solutions – like moving to the Bay Area – are already taken and priced in. If you want to win, you’re going to need to do something different and be right about it.