Oks argues that when firms cannot fire mid-career employees, redeploying existing workers into adjacent businesses becomes the only viable growth path. He cites Toray's pivot from mature textiles into carbon fiber — leveraging the same polymer-spinning engineers — as the canonical example of diversification driven by fixed labor rather than strategic whim.
Oks contends that the keiretsu system of cross-shareholdings and main-bank lending replaces the disciplining role of public equity markets, allowing diversified firms to allocate capital internally without paying a conglomerate discount. Because financing flows through relationship banks rather than dispersed shareholders demanding pure-play focus, sprawling business portfolios remain viable.
The editorial frames Prahalad and Hamel's focus-and-divest doctrine as a four-decade Western consensus that misread structural conditions as universal truths. Japanese firms ignored the advice and kept working by operational metrics, suggesting the conglomerate discount is an artifact of U.S. factor markets rather than evidence that diversification destroys value everywhere.
A blog post by David Oks titled *Why Japanese companies do so many different things* hit 615 points on Hacker News, surfacing a question Western MBAs have spent four decades trying to answer with the same wrong answer: why does Hitachi sell both nuclear reactors and rice cookers? Why does Yamaha make motorcycles and concert grand pianos? Why does Mitsubishi operate in shipbuilding, banking, beer, real estate, and aerospace simultaneously?
The Western consensus since the 1980s — popularized by Prahalad and Hamel's *The Core Competence of the Corporation* — has been that conglomerates destroy value. Diversified firms trade at a "conglomerate discount" of roughly 13-15% versus pure-play comparables in U.S. markets. Activist investors have spent decades dismembering GE, ITT, and Tyco. The orthodoxy: focus, divest, return capital to shareholders, let the market reallocate.
Japanese firms ignored this advice and, by most operational metrics, kept working anyway. Oks's piece walks through why — and the answer isn't culture or sentimentality. It's incentive structure all the way down.
The core argument is that Japanese corporate diversification is a rational response to factor markets that don't work the way American factor markets work. Three structural features do the heavy lifting.
First, lifetime employment. In a system where firing mid-career employees is socially and legally costly, a firm that wants to pivot has only one option: redeploy existing workers into new lines of business. If your labor force is fixed in size, the only growth axis left is to find new things for them to do. When Toray's textile business matured in the 1970s, Toray didn't fire textile engineers — it sent them into carbon fiber, where the same polymer-spinning expertise turned out to underpin a business that now supplies Boeing's 787 fuselage.
Second, bank-centric capital allocation. The keiretsu system — networks of firms organized around a main bank — meant capital didn't flow through public equity markets demanding quarterly focus. It flowed through long-term banking relationships that valued firm survival over portfolio purity. A bank lending to Mitsubishi Heavy doesn't care if MHI is in shipbuilding or air conditioners as long as the consolidated entity services its debt. This is the inverse of the U.S. setup, where index funds and activists punish diversification by repricing the stock.
Third, weak external labor markets for senior talent. In the U.S., a frustrated VP can quit Google on Monday and start at Meta on Tuesday with a higher comp package. In Japan, mid-career hiring is thin — your career capital is locked inside one firm. That makes internal job markets matter enormously, and internal job markets work best when the firm has many businesses to rotate people through. A salaryman's career is essentially a 35-year rotation through whatever the company happens to be doing — so the company needs to be doing many things.
The deeper insight, which Oks gestures at and the HN thread sharpens, is that "focus on core competency" was never a universal law of business — it was a contingent recommendation that assumed American factor markets. Pull out the assumptions of fluid labor, public equity primacy, and hostile takeovers, and the math flips. Hitachi's reactor division and rice cooker division aren't competing for the same capital pool the way GE Aerospace and GE Appliances were; they're sharing a labor pool, a R&D culture, and a banking relationship.
There's also a less-discussed angle: option value in long-horizon manufacturing. Sony spent decades as a consumer electronics company before pivoting hard into image sensors and entertainment IP — businesses that now produce the bulk of its profit. If you'd told Sony in 2005 to "focus on core competency," you'd have told it to double down on the dying Walkman/TV business and miss the CMOS sensor and PlayStation franchises entirely. The conglomerate structure is, in part, a hedge against being wrong about what your core competency actually is.
If you're a senior engineer or eng leader, the lesson isn't "go build a conglomerate." It's more uncomfortable than that: the management orthodoxy you absorbed by osmosis — focus, simplify, do one thing well — is downstream of capital-market structure, not engineering truth.
A few practical translations:
Internal mobility is a hidden moat. Companies that can rotate engineers across product lines without forcing them to quit and re-interview are running the Japanese playbook in miniature. Amazon and Google do this better than most U.S. firms; the rest watch their senior people walk because the only path to a new domain is a new employer. If you're building an eng org, your internal transfer policy is doing more work than your hiring funnel.
Adjacent diversification beats focus when the adjacent skill is real. Yamaha making both pianos and motorcycles sounds absurd until you realize the shared competency is precision metal casting and tuning — the same factory expertise that makes a tuning fork makes a carburetor. The right question isn't "are these businesses related?" but "do they share a non-obvious capability?" For software, the equivalent is: a payments team and a fraud team look unrelated until you notice they share a real-time event-processing platform. Don't optimize the org chart; optimize the substrate.
The "conglomerate discount" is partly a measurement artifact. Public-market investors discount diversified firms because they prefer to construct their own portfolios. But for private firms, founder-led businesses, and any organization not optimizing for stock multiple, that discount doesn't apply. If you're not going public, the spreadsheet that tells you to spin out the side business may be solving the wrong optimization problem.
The Japanese model is under genuine strain — millennial workers want career mobility, activist investors (notably Elliott Management) have started forcing Japanese firms to unwind cross-shareholdings, and Toshiba's recent breakup suggests the conglomerate-discount logic is finally biting. But the broader point survives: corporate structure is a function of factor markets, and as remote work, weakened non-competes, and AI-driven productivity reshape the U.S. labor market, the assumptions underlying "focus on your core competency" are themselves up for grabs. The next decade of org design may look less like the 1990s GE breakup and more like a 1970s Japanese trading house — if the labor side of the equation keeps moving the way it's been moving.
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.