Innovation in Real Places: Strategies for Prosperity in an Unforgiving World

By Dan Breznitz

10 min read

Why does it matter? Because the way your city is trying to grow is almost certainly making things worse for everyone who isn't already winning.

In 1903, silver was discovered near Cobalt, Ontario. Within a decade, millions had traveled north by train to one of the richest boomtowns on the continent. Today Cobalt has 1,100 people, one grocery store, and a pub. The train doesn't stop there anymore.

Three hundred miles south, Toronto never touched the silver directly. It built the stock exchanges, trading floors, and financing infrastructure the boom demanded — and became a global financial hub. Same boom. Completely different futures.

That gap is this book's entire argument in miniature. Most communities chasing prosperity are playing Cobalt's game: courting startups, idolizing venture capital, dreaming of becoming the next Silicon Valley. Dan Breznitz is here to tell you why that strategy almost always fails — and why the regions that actually create broadly shared prosperity are doing something far less glamorous, far more interesting, and almost universally ignored.

Innovation Is Not About Inventing Things — It's About Everything That Happens After

Dan Breznitz walked into a bicycle shop with his children to buy bikes. The design hadn't changed in a century: diamond frame, two wheels, pedals, chain, gears, brakes. Whatever innovation was possible seemed to have already happened.

Look at any bike in that shop and you'll find the same names: Shimano on the gears, probably Giant on the frame. Shimano is a Japanese company from Sakai City in Osaka. Giant is from Taichung, Taiwan. Neither looks like a hotbed of innovation. Both came to dominate a global industry through something much harder to see than invention.

Shimano, founded in 1921, spotted an opening after the oil crisis of the 1970s drove a global bicycle boom. It poured decades of engineering into gear and power transmission systems, thousands of incremental improvements, until no serious manufacturer could source gears elsewhere. Giant's route was less obvious. Founded in 1972, it grew to supply Schwinn, America's dominant bicycle brand, with 75% of its production. When Schwinn moved manufacturing to China in 1985 (a decision that led to Schwinn's bankruptcy in 1992), Giant faced collapse. Instead, it partnered with Taiwan's Industrial Technology Research Institute, the same body that had built the country's semiconductor industry, and developed carbon-fiber frames: stronger, lighter, and suddenly appealing to riders who had never owned a bike before. Giant now generates $2 billion in annual revenue.

Think of any global industry as four stages. Stage 1 is the part everyone talks about: the original invention, the startup, the VC-funded breakthrough. Stages 2, 3, and 4 cover design and production engineering, incremental improvement, and manufacturing at scale. Giant built its empire in Stage 3. Shimano too. The jobs, the wages, the community wealth, the pandemic resilience that comes from domestic manufacturing capacity — most of it came from stages that receive almost no attention from policymakers, almost no coverage in business media, and almost no money from the funds that claim to invest in innovation.

Most 'Innovation Districts' Are Just Talent Pipelines for Silicon Valley

Atlanta's "Silicon Peach" moment — the city's bid to become the South's answer to Silicon Valley — captures both the ambition and the delusion. The region had world-class engineering talent, a defense-industry origin story, Fortune 500 customers, and companies that genuinely dominated their markets: Scientific Atlanta in cable equipment, Hayes in modems, DCA's IRMA board among the top five best-selling PC products of the 1980s.

Then every winner collapsed or got absorbed. DCA and Hayes went bankrupt when the internet era arrived. MSA, once the world's largest software applications company, couldn't pivot from mainframes to PCs and sold for barely more than a single year's revenue. Scientific Atlanta was acquired by Cisco in 2006 and faded away. The number of large tech companies in Georgia dropped by more than half in six years. No Atlanta startup has since grown large enough to list on NASDAQ.

The structural reason was visible in the data. Breznitz and Mollie Taylor mapped the social networks of Atlanta's tech companies (the actual ties between founders, investors, engineers, customers, and peers). The finding was unambiguous: every company was isolated from its Atlanta neighbors. They interrogated the data every way they could, looking for something more encouraging. It wasn't there. Atlanta's entrepreneurs always walked alone.

The city wasn't short of talented people. What it lacked was the interlocking web of relationships — investors who know the engineers who know the customers — that turns individual companies into a community. Atlanta startups solved this the only way available: they wove themselves into Silicon Valley or Boston networks remotely. And that's where the trap closes. Every connection a founder builds in Menlo Park is one not built in Atlanta. Every hour spent cultivating a distant VC is time not spent on local ecosystem-building. The company becomes embedded elsewhere. Then it moves there.

The numbers accelerate this. Three-quarters of all US venture capital flows to California, New York, or Massachusetts. Silicon Valley alone draws more than the next four destinations combined. Ohio receives under one percent. When a Menlo Park VC funds your Atlanta startup, they're not investing in Atlanta. They're investing in a company that will be in their backyard soon enough. Of VC-backed Atlanta firms tracked over eight years, forty percent had left within three years of their first funding round.

That's the feeder cluster trap: regions build the companies; dominant hubs harvest them.

Even When Your Startup Ecosystem Works, the Financial Returns Flow Somewhere Else

Venture capital exports gains even when it wins. Two Canadian companies show how the structure determines who keeps them.

RIM bootstrapped with local investors; Canadian underwriters led its IPO, and profits stayed in Waterloo. Shopify took the VC path, and by IPO, American investors held over 42% of the company — shares that were then flipped to Canadian buyers at a premium. Same country, same sector, same era. Opposite direction of wealth transfer.

The reason is structural. A venture fund has roughly ten years to return outsized profits to its institutional backers. Partners collect 2-3% annually on the fund regardless of performance, and 20% of gains on exit. Their incentive is financial exit speed. One prominent Israeli VC told Breznitz that IPOs were "a pain in the neck" — he'd rather sell companies quickly to larger corporations and move on. That preference isn't a flaw; it's what the model requires.

Israel is where this plays out at scale. In 1968, the country had 886 R&D workers with academic training and the lowest business R&D investment in the OECD. By the late 1990s, it had more NASDAQ-listed companies per capita than any other nation. A genuine miracle, except that over 95% of the VC capital was foreign, and profits left with every exit. Productivity in every other sector stagnated or fell. Israel moved from the second-most-egalitarian society in the West to the second-most-unequal. One in five households now lives below the poverty line.

Breznitz co-authored the 2009 bestseller Start-Up Nation, which celebrated Israel as a model for other economies to imitate. He turns it here into his sharpest cautionary tale: the most successful innovation economy of the past fifty years enriched a narrow slice of its population while leaving the rest behind — not through failure, but through the very logic of the financial structure that powered it.

The Three Underrated Stages of Innovation Are Where Broadly Shared Prosperity Actually Lives

If venture capital siphons value to a handful of cities and old manufacturing can't return, what is everywhere else actually supposed to build?

The question contains a false binary, one that causes policymakers everywhere to walk past three forms of innovation. Stages 2, 3, and 4 (design and prototyping, incremental product improvement, and production) have created more broadly shared prosperity than all the VC-backed startups combined. They just don't get keynotes.

Taiwan is the proof. It built one of the twentieth century's most striking economic transformations through Stage 3: making existing products and technologies better, cheaper, and more reliable. In 1973, Taiwan's situation was precarious. Its textile and electronics exports were being undercut by cheaper competitors, key markets were closing off, and diplomatic isolation was deepening. The government targeted semiconductors and created a research institute called ITRI.

What followed would embarrass any startup pitch deck. For nearly two years, ITRI's engineers tried and failed to find any semiconductor company willing to share its fabrication technology. The partner they finally landed was RCA, a company exiting semiconductors entirely, willing to license a 7-micron process the rest of the industry had already left behind. The world frontier was 2 microns. Taiwan was starting four technology generations back.

Forty engineers spent nearly a year at RCA's US facilities mastering a process their host was abandoning. By 1979, the Taiwanese team was outperforming RCA on yield rates (the share of working chips in each production batch). The government then muscled a reluctant coalition of local companies into funding a commercial venture: United Microelectronics Corporation, which later spun off TSMC and MediaTek. The entire public investment, from ITRI's research budget through UMC's founding capital, came to roughly $35 million over less than a decade.

TSMC is now arguably the most strategically critical company on the planet. Taiwanese fabs produce the chips in every smartphone, fighter jet, and AI data center in use today. That position was built on technology RCA was discarding and forty engineers who learned to run it better than RCA ever did. Taiwan didn't invent semiconductors. It learned to make them better than anyone else and built an economy around that expertise.

What Taiwan built looks nothing like Silicon Valley prosperity: engineers, technicians, supply chain managers, machinists. Careers requiring real skill, not elite degrees. When Apple later tried to build an assembly plant in Austin, Texas, it couldn't even source the right screws. Decades of neglecting Stages 3 and 4 had dissolved the ecosystem those stages depend on.

Shenzhen followed a different path to the same conclusion. In 1979 it was a fishing town of 314,000 people with eight kilometers of paved road. Local officials didn't recruit star companies; they subsidized the suppliers and component-makers who would make those companies possible. By 2017, Shenzhen had produced Huawei, Tencent, DJI, and BYD, none of it built around prominent research universities. In Shenzhen, a missing component means a phone call to a neighbor, not a six-week import wait. The ecosystem, not any individual firm, was the asset.

The world runs on Taiwanese chips, Shenzhen supply chains, and the craft knowledge of Italian shoemakers — not on startup keynotes. The broadly shared prosperity that most regions want lives in these stages. And Taiwan got there starting with $35 million and technology a company was throwing away.

Don't Copy What Worked Elsewhere. First Understand Why It Worked There.

In 1983, Taiwan's Ministry of Finance issued regulations to create a venture capital industry. Officials had studied the American model carefully — they knew the structure, the terminology, the fund mechanics. What they built shared none of it except the name, and that's exactly why it worked.

American VC is a limited-time fund where passive investors supply capital and general partners control every investment decision, optimizing for rapid financial exit within ten years. Taiwan inverted this entirely. Investors created "paper companies" instead of funds and kept direct control of investment decisions. Companies had to show years of positive cash flow before listing on the Taiwan Stock Exchange. Taiwanese VCs couldn't chase the same exit logic even if they'd wanted to.

Capital stayed patient and local, backing the contract manufacturers and design-for-hire companies that Taiwan's economy ran on, not sprint-for-exit startups. Taiwan's officials started with a blunt question: how do we fund the kinds of companies we actually have? The American VC model had been designed to answer something else. Taiwan didn't fail to copy it. It rebuilt the machinery for different conditions and produced one of the world's most successful venture industries.

Copy the name without the diagnosis and you get a machine built for someone else's ecosystem.

Canada is the sharpest example of what goes wrong. Three universities in the world's top fifty. The highest public R&D investment among English-speaking nations. And since 2007, a measurable inverse relationship between government innovation spending and private business investment — the more taxpayers spend, the less industry invests.

Breznitz's verdict is direct: Canada has built an exceptional invention policy and keeps calling it innovation policy. They are different activities requiring different tools, and no increase in the dosage of the wrong medicine produces the right result.

The System Is Structurally Rigged Against Local Prosperity — Here's How to Navigate It Anyway

In May 2013, executives at the Timken Company of Canton, Ohio learned their shareholders were voting to break the company apart. Timken made high-end steel and precision bearings, including two-meter components for wind turbines, and had operated for nearly a century. MIT's Making in America study had just named it a model of regional development: 20,000 employees, deep community roots, exactly what a region building durable prosperity should want.

The attack was led by the California State Teachers Retirement System.

CalSTRS, the second-largest public pension fund in the country, had for decades been cited as the canonical example of patient, long-term capital. Investors demanded a breakup into "pure play" companies, arguing that making both steel and bearings was suppressing the stock price. They won.

The diagnosis: the problem was not short-termist hedge funds hijacking an otherwise sound system. CalSTRS managers acted completely within the law, doing exactly what regulations required. They maximized the utility function they were paid to uphold. The comfortable Star Wars narrative (CalSTRS as the Rebel Alliance, hedge funds as the Empire, the villain obvious and the hero's virtue assumed) was a story that kept everyone from seeing the real problem. Finance itself, including every regulation built around it, had become a structural obstacle to local growth. The odds it would change were, in Breznitz's view, close to zero.

The same verdict runs through intellectual property and data. Canada publicly funded the research that created modern AI. By the time the government launched its AI institutes in 2017, Microsoft held 1,030 machine-learning patent applications from the prior decade; all Canadian institutions combined held 48. The investment in science was real; the failure to treat IP as an extension of innovation policy meant the returns would accrue in Redmond, not Toronto.

Strategic realism is not despair; it is the prerequisite for action. The tools exist: conditional grants that let companies grow before financialization forces an exit, patent pools that make troll attacks expensive enough to deter, data clauses that keep locally generated information accessible to local innovators. But a policy menu is not a posture. A community that has internalized this doesn't wait for finance or IP law to be reformed. It assumes those systems will keep working exactly as designed (for someone else) and negotiates accordingly: demanding clawback provisions before public money flows to private firms, treating local IP as a public asset worth defending, asking what the exit terms are before celebrating a new foreign investment.

The moment an incumbent resorts to rent-seeking instead of competing, disruption is cheapest. The system is rigged — but understanding exactly how it's rigged is already an advantage.

The Best Time to Disrupt Is When Everyone Thinks the Game Is Already Over

The bleakest moment in any industry — when incumbents stop building and start extracting, when financial engineering replaces actual engineering, when monopoly protection substitutes for competition — is, counterintuitively, when you have the most room to move. The moment of peak stagnation is precisely when disruption is most possible, as Taiwan proved by licensing a process the US had already declared obsolete. They did it with a recycled technology license and $35 million. The question was never whether the global system would cooperate. It wouldn't, and it didn't. The question was whether anyone was clear-eyed enough about what stage of innovation they could credibly own, and stubborn enough to build the ecosystem that stage required.

The communities that lose aren't the ones with the fewest resources. They're the ones that convinced themselves the choice had already been made for them. It hasn't. But you do have to decide what kind of place you're actually trying to build — and then refuse to borrow someone else's answer.