The frontier stack thesis argues that semiconductors, robotics, space, and critical materials form one asset class, and that the cleanest way to hold it is through companies whose business is the frontier itself, not through diversified giants that merely touch it. That argument is only as good as the boundary it draws. This is the boundary: the rule that decides which companies are in the universe Robotnik tracks, and which are out.
The rule is a single test, applied by judgment and recorded with its reasoning. It uses no revenue threshold, and the next section explains why a number would be worse, not better. The universe it produces is a curated set of just under two hundred companies across the four sectors, actively maintained rather than fixed.
The test: core business, not incidental exposure
A company is in the universe only if its core business is the physical frontier stack value chain. It is out if the frontier is incidental to it, a side activity inside a company whose centre of gravity lies elsewhere.
That puts the diversified giants out, whatever their frontier work. A Big-Tech platform, a general-software vendor, an industrial conglomerate, a diversified defence prime, an energy major, a carmaker: each may do real frontier work, but for each the frontier is a fraction of something much larger, and the business as a whole is not a frontier stack business. Nvidia is in and Alphabet is out for this reason, and the reason is the shape of the business, not the size of the company.
Why there is no revenue threshold
A revenue floor would look more rigorous than a judgment, and it would be worse on three counts. It is arbitrary: there is no principled reason a company just above a chosen percentage belongs and one just below it does not, and the line would sit wherever it was first drawn. It is gameable, because how revenue is divided into segments is a reporting choice a company controls. And it measures the wrong thing, since a high frontier share can still describe a small division of a company that is plainly not a frontier business, while the question that matters is whether the frontier is what the company is for.
Serious classification is judgment applied consistently and shown its working, not a number that outsources the decision to a threshold. The rigour here is in the consistency of the test and the worked record of every ruling, the names taken in and the names left out, each with its reason. That record is what makes the boundary testable. Anyone can check a decision against the rule and its precedents, and argue it.
Dominance, not criticality
The test turns on what a company mostly does, not on how important its products are to everyone else. Those two things come apart, and the rule follows dominance.
A company can make something the frontier cannot do without and still be out, if that product is a minority of a larger non-frontier business. Industrial-gas suppliers such as Linde and Air Products are the clearest case. Their gases are essential to chip fabrication, but the semiconductor business is a minority of a broad industrial company, so they are out of the universe.
This is deliberate, and it is where the membership rule and the control-point lens do different jobs. A control point is a place in the supply chain where a critical input is concentrated. It is a fact about the input, not about who supplies it. The universe is the set of companies whose business is the frontier. It is a fact about the company. The two often diverge: a chokepoint can be held by a company the universe excludes, because the company is not a frontier stack business even though its product is a frontier stack dependency. Robotnik tracks those critical inputs in its dependency and bottleneck analysis regardless of who makes them, and admits their makers to the equity universe only when the frontier is their core business.
The four sectors
Every member is placed in one of four sectors by its core business: semiconductors, robotics, space, or materials and inputs. The classification is re-derived for each name from what the company actually does, not taken from the sector labels that data providers attach, which are often wrong at the edges. One company, one sector. A name whose business spans two is placed by where its centre of gravity sits, and the reasoning recorded so the call can be checked.
Defence: the conglomerate test
Defence is the case that most often looks like a contradiction, so the rule states it plainly. A diversified defence prime is out, even where it runs a real space or robotics division, because the company as a whole is a defence conglomerate and the frontier work is one part among many. Raytheon, Lockheed Martin, Northrop Grumman, and General Dynamics are out. A pure-play frontier-defence company is in, because the frontier is its whole business. Kratos and AeroVironment are in. The line is diversified against pure-play, the same test as everywhere else, and removing the primes is what lets Space stand as a pure-play sector instead of a wing of the defence industry.
Energy transition: the materials, not the rest
The energy transition sits next to the frontier stack and overlaps it at one point: materials. Battery and magnet materials are in, under materials and inputs, because they feed the same critical-materials chain the stack depends on, and because the commodities index already counts energy-transition demand when it weights those materials. Lithium and rare-earth producers such as Albemarle, SQM, MP Materials, and Lynas are in. The rest of the energy transition is out: solar-panel makers, nuclear and uranium names, grid and general-electrical companies, and electric-vehicle assemblers. First Solar, Oklo, and Tesla are out. The universe takes the upstream materials the frontier shares and leaves generation, grid, and vehicles to their own sector.
Companies before revenue
Much of the frontier is pre-commercial, especially in space and advanced robotics, so a rule that turned on revenue would exclude the most frontier-native companies of all. The core-business test does not. A company with little or no revenue is admitted on the same basis as any other, on what its business is, not on what it earns yet. Rocket Lab, AST SpaceMobile, and Intuitive Machines were members while still pre-profit.
These names carry a lifecycle state, running from private through the stages of a public listing to delisting or acquisition. A company that has not yet listed can be shown but is walled off from the indices: it carries no index weight, and a private funding-round valuation is never turned into a market capitalisation. A valuation set by a private round is not a market price, and treating it as one would fabricate a number the market has not made. Robotnik records the valuation as what it is and keeps it out of any figure that implies a traded price.
A managed universe
The universe is managed, not fixed. It is reviewed as the evidence changes, and names enter, leave, or move between sectors accordingly, each disposition recorded as in, out, contested, or readmitted. Every company keeps a stable identifier through its whole life, so that a name changing hands, listing, or spinning off updates cleanly without corrupting its history or counting it twice. Mergers, spin-offs, initial public offerings, and delistings are handled as transitions on that identifier, not as deletions and fresh additions.
Where a company's filings do not make its core business obvious, the call is a judgment, made per name and recorded with its reasoning. A name that cannot be assessed with confidence is held out until it can be, not guessed in. The same discipline runs through the whole registry: flag what is uncertain, never fabricate it.
Where this rule is weakest
A judgment-based rule is only as good as the judgments behind it. Its weakest points are the genuine borderlines: a company shifting from a diversified base toward a frontier core, or the reverse; a name whose two businesses are close to balanced; a sector assignment a reasonable analyst would make differently. These are marked as contested rather than smoothed over, and they are where the review does its real work. The rule will get individual calls wrong. The commitment is to make each call in the open, against a stated test, so that a mistake is visible and correctable instead of buried inside a number.
Membership and the value chain
Membership decides whether a company is in the universe. It does not describe what the company does within it. That is the work of the value-chain taxonomy, which places every member at its position in the stack, from the upstream end of design and materials through to deployment and the services around it. Membership is the boundary; the value chain is the map inside it. The two are set out separately because they answer different questions.
Frequently asked questions
How does Robotnik decide which companies are in its universe?
By one test: whether the frontier stack is the company's core business. A company is in if semiconductors, robotics, space, or critical materials is what it is fundamentally for, and out if that work is incidental to a larger business elsewhere. The decision is a judgment, applied consistently and recorded with its reasoning.
Why is there no revenue threshold for inclusion?
Because a threshold is arbitrary, gameable, and measures the wrong thing. There is no principled cut-off on a frontier-revenue share, segment reporting can be arranged, and a high frontier share can still belong to a minor division. A core-business test applied consistently, with every ruling published, is more rigorous than a false-precise number.
Why is Nvidia in the universe but Alphabet out?
Because membership turns on the shape of the business, not its size. The frontier is effectively all of Nvidia's business and a fraction of a per cent of Alphabet's, whose revenue is overwhelmingly advertising. Nvidia is a frontier stack company; Alphabet is a conglomerate that does some frontier work.
Are defence companies in the universe?
Diversified defence primes are not, because they are defence conglomerates for which frontier work is one division among many. Pure-play frontier-defence companies, whose whole business is the frontier, are. The test is the same as everywhere: diversified out, pure-play in.
Are solar, nuclear, and electric-vehicle companies included?
No. Of the energy transition, only the upstream materials are in, the battery and magnet materials under materials and inputs, because the frontier shares that critical-materials chain. Solar-panel makers, nuclear and uranium names, grid companies, and vehicle assemblers are out.
How are pre-IPO and pre-revenue companies handled?
They are admitted on the same core-business test, since much of the frontier is pre-commercial. A company that has not yet listed can be shown but carries no index weight, and a private valuation is never converted into a market capitalisation, because it is not a market price.
How often does the universe change?
It is managed rather than rebuilt on a fixed schedule. Companies enter, leave, or move between sectors as the evidence changes, and corporate actions such as mergers, spin-offs, and listings are handled as transitions on a stable company identifier, so the history stays intact.