Information
Leads to
Innovation.

Most founders read a cap table as a map of power — who owns what, who controls the vote. That is the smaller half of the story. A cap table is also a tax document, and the choices made on the first day, before there is revenue or even a logo, quietly decide the bill a founder pays years later when the company finally works.

Start with the 83(b) election. When you take founder stock subject to vesting, the IRS gives you thirty days — thirty, not sixty, not “about a month” — to file a one-page election asking to be taxed now, on a value near zero, rather than later, as the stock vests into real money. Miss the window and every tranche that vests becomes ordinary income at whatever the company is then worth. Founders have built six-figure tax bills on shares they could not sell, simply by letting thirty days pass.

Then there is the entity. Everyone forms an LLC because someone said it was simpler. For a venture-track company, the LLC is often the most expensive convenience in the building. The largest tax break available to founders — the qualified small business stock exclusion under Section 1202 — requires C-corporation stock, issued at the right time and held for five years. Begin as an LLC, convert late, and you restart that clock against the very gain you most wanted to protect.

Even the instruments you raise on carry tax weight. A SAFE is not stock; it converts into stock at a future round, and the holding period that matters often does not begin until conversion. Priced rounds, convertible notes, and SAFEs each start different clocks and create different basis. None of it appears in the ownership column, and all of it arrives at exit.

The lesson is not that founders should become tax lawyers. It is that the cheapest hour you will ever spend is the one with someone who reads a cap table the way an auditor does — before the structure hardens, while a stamp can still fix it. We invest early in part because that is when good structure is still free. After the Series B, it is surgery.

There is a line in the Internal Revenue Code that can turn a founder’s exit from a taxable event into a tax-free one, and most people building companies have never heard its name. Section 1202 — qualified small business stock — lets a founder exclude the greater of ten million dollars or ten times their basis in gain from federal tax. Not defer. Exclude.

The requirements are specific, which is exactly why founders miss them. The stock must be in a domestic C-corporation. The company’s gross assets must be fifty million dollars or less at the moment the stock is issued. The business must be a qualified trade — which rules out most finance, law, consulting, and health services, and includes most software and product companies. And the stock must be held for five years.

Each requirement is a trap for the unprepared. The C-corporation rule punishes the founder who started as an LLC and converted late. The fifty-million-dollar ceiling is measured at issuance, so shares issued early — when the company is small — qualify, while shares issued after a large round may not. The five-year clock rewards the founder who took stock at formation and punishes the one who papered their equity two years in.

What almost no one uses is stacking. The ten-million-dollar exclusion runs per taxpayer, per company — so a founder who gifts qualifying stock to a non-grantor trust or to family members can multiply the exclusion across several taxpayers, legally, when it is done early and cleanly. A single exit can shelter several times the base amount across a thoughtfully built structure. The founders who capture it did the work before the company was worth anything, when giving away stock felt like giving away nothing.

Section 1202 is the rare provision that rewards founders for being early and organized — the same traits that build companies. It is written in a language most cap tables never learn to speak.

Founders obsess over one number on a term sheet — the valuation — and sign away the four that decide what they keep. The valuation is the headline. The terms are the contract. A higher number with hostile terms is routinely worth less than a lower number with clean ones, and the people who reliably know the difference are the ones who have watched the exits.

Begin with the liquidation preference. A “1x non-participating” preference means an investor takes their money back or their ownership percentage, whichever is greater — standard and fair. A “participating” preference means they take their money back and then share the rest as if they had not. Stack a participating preference across a few rounds and a founder can sell for nine figures and watch most of it walk past them to the preference stack. The valuation looked generous. The waterfall was not.

Then anti-dilution. “Weighted-average” protection nudges an investor’s conversion price if a later round prices lower — a shock absorber. “Full ratchet” reprices their entire stake as though they had invested at the lower price, transferring ownership from founders and employees to the earlier investor wholesale. One phrase in a defined term decides who absorbs the pain of a down round.

The quiet clauses matter too. The option-pool expansion taken before the round rather than after dilutes only the founders while wearing the costume of good governance. Pro-rata rights decide who keeps their percentage in the rounds that count. Board composition decides who controls the company once the signing dinner ends.

None of this means founders should fear term sheets. It means reading them the way the other side does — as a machine for allocating money and control under conditions no one is picturing on signing day. We tell our founders the same thing every time: negotiate the terms, not the headline.

A founder issues a token to fund a network, calls it a utility, and decides it is not a security because it does something. The SEC has spent the better part of a decade explaining, one enforcement action at a time, that doing something is not the test. The test is older than crypto, older than the internet, and it does not bend.

It comes from a 1946 case about orange groves. Under the Howey test, an investment contract — and therefore a security — exists when people invest money in a common enterprise expecting profit from the efforts of others. Almost every token sold to fund a team building a network fits that sentence cleanly. The label on the token does not matter. The pitch does. A founder who sells tokens by promising that their work will make those tokens worth more has just described the fourth prong of Howey in their own deck.

This is why the functional-token defense so often fails. A token that genuinely runs a live network, bought to use rather than to flip, drifts away from the definition. A token sold before the network exists, to people betting the team will build value, sits squarely inside it — however the whitepaper is worded.

The paths that work are the boring ones. Sell to accredited investors under a Regulation D exemption. Use Regulation A+ for a broader raise with real disclosure. Structure the instrument as what it is and accept the registration or exemption that follows, rather than litigating the question after the fact against an agency that has already won this argument many times.

Crypto founders are not wrong that the rules are unsettled at the edges. They are wrong about where the edges are. The center of the field — raising money from the public to fund a team’s efforts — has been marked since 1946. The founders who last treat securities law as architecture, not as an obstacle to route around.

For thirty years the binding constraint in technology was the chip. Faster processors, smaller transistors, more compute per dollar — the whole industry organized itself around an improving curve. That era is closing, and the new constraint is not the chip. It is the electricity to run it.

A modern AI cluster does not ask for a better processor so much as a substation. Training and serving large models draws power at a scale that looks less like a software company and more like a heavy-industrial plant. The chips exist; the megawatts to feed them, and the grid connections to deliver those megawatts, increasingly do not. The waiting line to plug a large load into the grid now stretches for years in the regions where data centers most want to sit.

This changes what it means to build a compute-heavy company. A founder can raise capital, buy accelerators, and still be unable to switch them on, because the limiting resource is a power contract and a place in line, not a purchase order. The companies winning the infrastructure race are increasingly the ones who secured energy first and solved the software second — the inverse of how the last generation was built.

For early-stage founders the implication is concrete. If your cost structure is dominated by inference, your real competitors are not other apps; they are everyone bidding for the same constrained power and the same scarce rack space. The edge goes to teams who treat energy as a first-class input — who know their cost per token at the level of the electricity bill, not just the API price.

Every platform shift eventually meets a physical limit the software people did not see coming. This one is the oldest limit in industry: who controls the power. Founders who understand that early will build differently, and raise differently, than the ones still assuming the only scarce thing is talent.