The Ten-Year Target: Long Bets and Institutional Patience

At the top of Traction’s Vision/Traction Organizer sits the ten-year target: one dated, measurable destination — Wickman’s small-company rendering of the BHAG, the Big Hairy Audacious Goal that Collins and Porras introduced in Built to Last and formalized in their 1996 HBR piece “Building Your Company’s Vision.” The specification matters more than the pedigree: it’s one target, not a paragraph of aspirations; it’s far enough out that you can’t backward-plan to it; and everyone in the company can recite it. Its job in the V/TO stack is to be the thing the three-year picture is a waypoint toward and the one-year plan navigates by.

For innovation strategy, though, the interesting question isn’t how to write a ten-year target — an afternoon’s work — but how to hold one, because a decade-scale bet must survive roughly forty quarterly earnings conversations, several budget crunches, and at least one leadership change, every one of which is an opportunity to quietly stop. The evidence says holding is worth engineering for: McKinsey Global Institute’s Corporate Horizon Index, tracking 615 US large- and mid-caps from 2001 to 2015, found firms managed for the long term delivered average revenue growth 47% higher and earnings growth 36% higher than their short-term peers. (The methodology has its critics — selection effects are hard to expunge — but the direction survives every re-cut I’ve seen.) So this post is a tour of decade-bets that paid, the machinery that let their owners stay committed, and the two opposite ways such bets die.

What a held decade-bet looks like

AWS is the canonical case, and the under-told part is the disclosure discipline. The infrastructure-as-product thinking dates to around 2003; S3 and EC2 launched in 2006; and Amazon then ran the business for nearly a decade without breaking out its numbers — until April 2015, when the first segment disclosure revealed AWS was already the company’s most profitable division. By 2024 it was a business with over $100B in annual revenue and roughly $40B in operating income. Bezos’s line — that Amazon is “willing to be misunderstood for long periods of time,” delivered at the 2011 shareholder meeting among other places — usually gets filed as a quote about courage. It’s actually a quote about information management: not disclosing AWS’s economics for nine years was a structural choice that denied the market the data it would have needed to demand the bet be resized.

Apple silicon shows the same shape in capability form: the PA Semi acquisition in April 2008 (about $278M) to the M1 announcement in November 2020 is a twelve-year arc — through A-series iPhone chips that each looked like an end in themselves — arriving at control of the Mac’s entire compute stack. The lesson for a ten-year target: the target was a capability position (“design our own silicon”), not a product, and every intermediate product paid rent on the way. A decade-bet that generates no interim value must survive on faith alone; one structured like Apple’s compounds evidence and revenue as it goes.

ASML’s EUV is the outer limit of the genre. Extreme-ultraviolet lithography went from research effort (mid-1990s) through first pre-production machine (2010) and first production model (2013) to volume-capable systems in the late 2010s — with ASML’s own account putting its EUV R&D above €6 billion across roughly seventeen years, inside an industry-wide effort estimated at two to three times that. The machinery of patience here was shared risk: in 2012 ASML sold equity stakes to its three biggest customers — Intel, Samsung, TSMC — who co-funded the R&D, converting the customers most exposed to the bet’s failure into co-owners of its continuation. TSMC itself is the same story at national scale: founded 1987 with the Taiwan government’s development fund holding nearly half the equity and Philips contributing capital plus technology, on Morris Chang’s then-heretical pure-play foundry thesis that took over a decade to look obviously right.

Toyota’s Prius compresses the pattern into corporate form: the G21 project chartered in 1993 as an open question about the car for the next century, an executive (Wada) doubling the fuel-efficiency target mid-project to force a technology leap rather than an optimization, hybrid approval in 1995, and launch in Japan in December 1997 — four years from charter, inside what became a twenty-year hybrid position no competitor matched until the market had already conceded it. And SpaceX’s reusability program — announced September 2011, mocked through four years of public crash landings, first successful ground landing December 2015, first re-flight of a recovered booster March 2017 — is the reminder that a decade-bet’s middle years look like failure by design, which is precisely why it needs protection from quarterly judgment.

The machinery of patience

Strip those stories and the same three mechanisms recur. Ownership structure: Google’s founders wrote dual-class voting into the 2004 IPO explicitly — the Founders’ Letter promises long-term bets and no earnings guidance in nearly the same breath — and that structure is what later absorbed DeepMind’s years of nine-figure losses (roughly £2B cumulative before its first profit in 2020, including a £1.5B debt waiver) without a shareholder revolt. Funding structure: metered but pre-committed — ASML’s customer co-investment, TSMC’s state anchor, IBM’s EBO class of money — so the bet’s continuation isn’t re-litigated from zero at every budget cycle. Measurement structure: decade-bets must be scored on milestone evidence, not P&L — the innovation accounting discipline, held for ten years instead of ten weeks. A ten-year target announced without all three is a press release with a long fuse.

The two deaths: too little commitment, and too much

Decade-bets die in opposite ways, and the famous corpses split cleanly.

Under-commitment is the Kodak and Nokia pattern — and the Kodak version deserves its accurate telling, which Scott Anthony’s HBR piece supplies: Kodak didn’t ignore digital. Steve Sasson built the first digital camera prototype there in 1975, and the company later invested billions in digital imaging. What Kodak never committed to was the business-model consequence — that photography’s future was sharing, not printing — so every digital investment was quietly shaped to protect the print franchise. Nokia ran the same play with hardware: touchscreen phone and tablet prototypes years before the iPhone, confirmed by its own former designers, shelved for fear of cannibalizing the handset lines. In both, the technology bet existed; the target — a stated, recited destination that would have forced the cannibalization question — did not. That’s what a ten-year target is for: it converts “should we disrupt ourselves?” from a battle refought yearly into a decision already made. (The fuller strategy autopsies of both companies are in the Nokia and Kodak product story.)

Over-commitment is the Concorde, which gave the failure mode its academic name: development costs ran on the order of twenty times the initial 1962 estimate, the economics were visibly unworkable years before launch, and two governments kept paying anyway — prestige and sunk cost doing the deciding — through twenty-seven years of service that never approached commercial viability. “Concorde fallacy” entered the literature via Dawkins and Carlisle in Nature in 1976 as the biologists’ term for escalation of commitment. The uncomfortable symmetry: patience and escalation feel identical from the inside. Both look like “staying the course under criticism.” The difference is falsifiability — AWS, ASML, and the Prius all carried testable interim milestones that kept passing; Concorde’s milestones kept failing and the funding continued. A ten-year target needs kill conditions the same way a quarterly rock does; the horizon changes the clock, not the epistemics.

The cheapest illustration that money and patience aren’t the same thing predates all of these: in 1903 the US War Department had granted Samuel Langley some $50,000 — around $2M today — for a heavier-than-air flying machine that twice crashed into the Potomac, while the Wright brothers spent roughly a thousand dollars of bicycle-shop revenue and flew that December. The Wrights’ advantage, as the Scientific American retrospective argues, was learning velocity — hundreds of glider iterations and a homemade wind tunnel against Langley’s two full-scale, all-or-nothing launches. A decade of runway buys nothing if the bet only generates evidence twice.

Writing yours

The mechanics, back at V/TO altitude. One target, dated, measurable, recitable — a position or capability (“our own silicon,” “every booster flies twice”) rather than a product, so interim products can pay rent along the way. Cannibalization answered in the text: if the target comes true, name what it kills — the sentence Kodak and Nokia never wrote down. The patience machinery named at declaration: whose money is pre-committed, what structure protects the bet from quarterly re-litigation, and what evidence cadence it reports on. Kill conditions, written the day it’s announced, so your successors can tell patience from Concorde. Then hand it down the stack: the three-year picture paints the next waypoint toward it, and the one-year plan turns the waypoint into rocks. That’s the whole V/TO, run on innovation — vision at the top, traction underneath, and no hallucination in between.