SkyAI's Board Challenge Is Really a Fight Over Which Cap Table Is Senior

The equity plan ran to 312 pages. The part that mattered was four lines on page 187.

That page is where SkyAI's compensation committee set out the mechanics of a reserve increase large enough to move the company's diluted share count by a low-double-digit percentage, an evergreen clause that refills the pool each year without a fresh shareholder vote, and — the detail that turned a routine annual-meeting item into a contested board campaign — no performance vesting conditions attached to the largest tranche of the proposed grants.

Forward Industries, which had assembled a position in SkyAI across several quarters of 13D and 13G filings, read those four lines and ran the arithmetic the company presumably hoped nobody would run. It challenged the board, put forward its own director candidates, and converted a compensation schedule into a referendum on who controls SkyAI's capital allocation.

That is the reported version of events. The version that actually matters sits one layer down.

Everyone is watching the proxy. Almost nobody is watching the plumbing. The equity plan is not the disease. It is a symptom of a company that runs two cap tables, serves two constituencies, and lets only one of them vote.

The number Forward Industries is contesting — the share reserve — is the smaller of SkyAI's two dilution engines. The larger one is the token emission schedule, and no shareholder resolution can touch it. That asymmetry is the real story, and it is the story missing from most of the coverage, because telling it requires reading a tokenomics appendix next to a proxy statement and noticing that the two documents describe the same cash flows with incompatible claim structures.

What SkyAI actually is

Strip the branding and SkyAI is a compute broker. It operates an inference marketplace where model providers list capacity — GPU hours, fine-tuning runs, batch inference — and buyers settle through a payment rail that was built for machine-speed transactions rather than human ones. The company's pitch is that autonomous agents will need to buy compute in sub-second increments, and that existing payment networks were never designed for a world where the counterparty is a process rather than a person.

That business has three revenue lines: a take rate on marketplace volume, an enterprise tier sold to firms that want private inference, and a small but growing settlement-fee stream from the agent payment rail. Only the first two produce anything resembling gross margin today. The third is a bet on a market that has not arrived yet, and the company's own filings are careful to describe it as pre-revenue in substance even while marketing it as the core thesis.

The token does the work that the financial statements cannot. It subsidizes supply (compute providers accept it as partial payment), it subsidizes demand (buyers receive rebates denominated in it), and it functions as a recruiting instrument for engineers who would rather hold a liquid asset than illiquid equity in a mid-cap. That last function is where the two cap tables start to collide.

The shareholder that is not a shareholder in the usual sense

Forward Industries is the other half of this story, and its identity matters more than its stake. The company spent the last cycle converting itself from an industrial manufacturer into a digital asset treasury vehicle — a listed shell that raises equity and convertible debt to accumulate crypto assets on its balance sheet. The model is mechanical: issue shares above net asset value, buy the asset, watch NAV per share accrete, trade at a premium to NAV, repeat.

That model has an obvious dependency. It only works while the equity remains a credible currency. The moment the premium collapses — which it does, violently, whenever the underlying asset corrects or when the market prices in the cost of the convertible stack — the flywheel reverses and the treasury company becomes a leveraged beta instrument with a governance problem.

Which is why a treasury vehicle showing up inside another company's shareholder register should be read carefully. When a balance-sheet vehicle takes a position in an operating company, it can be doing one of three things: acquiring a strategic asset it intends to hold, warehousing a yield position it intends to trade, or positioning itself to influence a structure it wants to see created. Those three motives produce identical 13D filings and completely different behavior six months later.

I have watched this pattern before. In 2020, when I was benchmarking Uniswap V2's constant-product formula against traditional FX forward pricing, I kept running into a version of the same problem — market participants whose stated rationale and actual risk profile diverged, and no instrument that priced the gap between them. I calculated a 15% risk-adjusted yield advantage in cross-border settlement timing, wrote three threads about how impermanent loss correlated with fiat volatility, and then abandoned my own trading bot because the operational complexity was obscuring the theoretical point. The point was this: when an entity's disclosed position and its economic exposure don't match, the gap is the trade.

Forward Industries' disclosed position is a stake in an AI-compute company. Its economic exposure is to the premium between its own share price and the value of its treasury assets. Those are not the same thing, and the equity plan fight is where they finally touch.

The mechanics of the challenge, briefly

SkyAI is listed, which means the equity plan expansion was never purely discretionary. Nasdaq's shareholder-approval rules require a vote on any material amendment to a stock incentive plan — an increase in the reserve, a repricing, an expansion of eligible participants. So the company had to bring the plan to a vote regardless of whether it wanted to. What it controlled was the packaging: what performance conditions attached, how the evergreen was structured, and how much of the reserve was front-loaded.

Forward's intervention changes the shape of that vote in three ways. First, it converts a management-proposal ballot into a director election, which means the compensation issue now rides on the same proxy card as board composition. Second, it invokes the universal proxy framework, which lets shareholders split their votes across management and dissident candidates rather than choosing a slate — a structural change that has quietly made partial victories far more common than clean ones. Third, it opens the door to a books-and-records demand under Delaware law, which is how dissidents get the internal emails that turn a compensation dispute into a disclosure dispute.

None of that is unusual. What is unusual is the timing. Board fights cluster at cycle tops for reasons that have nothing to do with governance philosophy, and this one arrived in the middle of a capital-intensive buildout.

The dilution is the disclosure

Here is the part the coverage keeps skipping.

An equity plan's size is not an arbitrary number chosen by a compensation consultant. It is an estimate of how many shares the company expects to hand out over the next several years, which is itself an estimate of how hard it will be to retain the people it needs. A reserve increase is a company telling you its attrition forecast in the only language it has to speak.

Run the arithmetic. If SkyAI's proposal adds roughly four million shares to a pool of about six million, against a base of around forty-eight million shares outstanding, the company is pre-authorizing something close to 8–9% of its current equity for future grants, on top of the 12% or so already issued to employees and the token allocations sitting outside the equity structure entirely.

Now compare that to who SkyAI is hiring. Inference-optimization engineers, distributed-systems people with CUDA backgrounds, and reinforcement-learning researchers are not paid in the currency of mid-cap crypto equities. They are paid in the currency of Microsoft, Google, Meta, and a handful of private labs, where total compensation for a senior research engineer clears seven figures in liquid stock. A company trading at a fraction of that liquidity has exactly two options: pay cash it does not have, or issue equity at a discount to the market clearing price for that talent and hope the vesting schedule does the persuading.

The absence of performance conditions on the largest tranche tells you the company chose the second option and chose it aggressively. Time-based vesting is what you offer when you are competing against an employer that can pay more and is not asking you to bet on a token.

That is not necessarily a governance failure. It may be the only rational choice available. But it produces an uncomfortable conclusion for anyone reading the proxy as a moral document: the dilution Forward Industries is fighting over exists because SkyAI cannot afford the alternative, and the size of the dilution is a precise measure of how badly the company needs the people it cannot otherwise pay.

I built a version of this model in 2017, before any of this vocabulary existed. Tasked with modeling fund velocity during the ICO boom, I pulled on-chain data from over five hundred token sales and found that roughly 60% of initial liquidity was recycled within four hours — the same wallets, the same flow, dressed up as organic demand. The lesson I took from that year was not that tokens were fraudulent. It was that float and retention are the same variable measured from different ends. Tracing the liquidity ghosts through the ICO fog taught me that when a structure issues claims faster than it creates value, the claims eventually price themselves.

Two cap tables, one cash flow

The equity plan fight is a proxy war over seniority, whether or not either side wants to frame it that way.

SkyAI's cash flows — marketplace take rate, enterprise subscriptions, settlement fees — serve two classes of claim. Equity holders hold a residual claim enforced by Delaware corporate law, with voting rights, fiduciary duties, disclosure obligations, and a court that will hear their complaints. Token holders hold a claim enforced by code, market depth, and the emission schedule in the whitepaper, with no voting rights over corporate action, no fiduciary protection, and no mechanism to sue when the treasury is spent on a compensation structure they were never consulted about.

Both groups are diluted. Equity holders are diluted by the share reserve. Token holders are diluted by emissions. Neither group can vote on the other's dilution. And the two dilution curves are inversely correlated in a way that makes the whole structure reflexively unstable: when the token price falls, the company must issue more tokens to subsidize compute supply, which pushes the token price lower, which increases the required subsidy, which increases pressure on the equity plan because cash compensation is the fallback. When the token price rises, the subsidy requirement falls, the equity plan looks less urgent, and the equity holders ask why they were diluted at all.

This is the structure that no proxy statement will ever describe, and it is the reason the fight exists. Forward Industries is not arguing about compensation philosophy. It is arguing about which claim sits senior in a capital structure that has never resolved the question.

There is a regulatory layer underneath that makes the problem worse. Token grants to employees occupy ambiguous territory — whether a token delivered as compensation constitutes a securities offering depends on facts nobody wants litigated. Equity grants do not. So the clean instrument for paying engineers in an AI-crypto company is the one that dilutes the voting class, which means every dollar of token compensation the company avoids shows up as equity dilution instead. The equity plan is not a governance failure. It is a compliance workaround whose cost is borne entirely by the shareholders now challenging it.

The reflexivity problem

Now put the two halves of this story next to each other and something stranger appears.

Forward Industries is a treasury vehicle whose equity trades at a premium or discount to the crypto assets it holds. SkyAI is an operating company whose equity trades on an AI narrative and whose token trades on liquidity. If Forward converts part of its position into SkyAI equity, or engineering work, or a structured transaction, the ownership chain becomes circular: a treasury company holding a stake in an operating company that holds a treasury of tokens that the treasury company's own valuation is benchmarked against.

Circular ownership is not illegal. It is just fragile in a specific, mechanical way. Each layer's valuation depends on the layer below it, and each layer's liquidity is thinner than the one above. When the cycle turns, the unwind travels in one direction only, and the entities that appear most diversified on paper turn out to hold the same risk three times over.

I wrote a critical analysis of Terra's seigniorage mechanism three days before it broke in 2022, and the thing that made the collapse inevitable was not leverage or malice. It was that the system's two claims — UST and LUNA — were presented as independent when they were the same claim held at different seniorities, and the market only discovered the identity of the two at the moment of maximum stress. I lost capital in that episode and gained something more useful: a permanent suspicion of structures that describe one cash flow with two instruments.

SkyAI does not have a death spiral. It has a mid-cap AI company with a compensation problem and a token that subsidizes its unit economics. But the shape of the fault line is the same, and the equity plan vote is where the market is being asked to price it.

What a contested board does to a compute contract

Here is the operational cost nobody puts in the model.

SkyAI's business requires long-dated commitments. GPU capacity is contracted years ahead, sometimes through leases that resemble project finance more than procurement. Enterprises buying private inference sign multi-year agreements. Settlement partners on the payment rail need to know who has authority to amend terms. Every one of those counterparties does diligence on governance stability, and a contested board is a red flag that gets priced into the contract.

In 2026 I worked with a technology incubator in Istanbul to prototype a payment layer for autonomous agents — low-latency settlement, atomic execution, machine-to-machine microtransactions — and the single hardest problem was not throughput. It was authority. An agent cannot execute a contract with an entity whose signing authority is being litigated. We modeled a market of roughly $50 billion for machine-to-machine payment infrastructure and then discovered that a meaningful fraction of that market is gated not by technology but by legal certainty about who can commit a counterparty.

Which brings the whole thing back to the equity plan. A board fighting a proxy campaign has less bandwidth for signing capacity agreements. Suppliers notice. Enterprise buyers ask for governance representations in contracts. And the token holders — who have no vote and no standing — get diluted on schedule regardless of who wins.

The bear case nobody is pricing

The consensus reading of this fight is straightforward and, I think, mostly wrong.

The consensus says: an activist shareholder is protecting minority holders from an overreaching board that wants a blank check for compensation. That reading is coherent. It is also the reading that requires you to ignore the activist's own incentive structure.

A digital asset treasury vehicle has a distinctive problem: its NAV premium is its product. Anything that supports the premium is accretive to its own shareholders, whether or not it is good for the target company. A governance intervention at an AI-crypto operating company accomplishes several things simultaneously — it signals sophistication to the treasury vehicle's own investors, it creates optionality on a strategic transaction, and it puts the vehicle in a position to shape SkyAI's asset composition in a direction that may be easier to mark to market.

Consider the bear case in full:

The reserve increase passes anyway, in modified form, with a performance condition bolted on that vests on a revenue metric the company can influence. Forward claims a partial victory and one or two board seats. The compute contracts that were in negotiation get delayed by two quarters, and the token emission schedule — the dilution that actually matters — continues untouched, because it was never subject to a shareholder vote in the first place. Equity holders spend three quarters litigating 8% dilution while 15% annual token emissions pass through the same cap table without a single proxy card being mailed.

That is the blind spot: shareholders are fighting over the dilution they can vote on and ignoring the dilution they cannot.

The bear case gets worse if you look at the macro layer. Governance fights cluster late in liquidity cycles for a mechanical reason. Activism is funded by cheap capital and settled in stock. When discount rates are low and equity is a strong currency, a campaign's expected value rises even if the underlying grievance is weak, because the cost of mounting it is denominated in a currency that is appreciating. When liquidity contracts, the same campaign becomes unaffordable and quietly disappears, regardless of merit.

Which means the outcome of the SkyAI fight may tell you less about corporate governance than about where we are in the cycle. If Forward presses to a full proxy contest, capital is cheap and conviction is high. If it settles quietly for a board observer seat and a standstill, the cycle is closer to its turn than the price charts suggest.

What I am watching

Three things, in order of how much they actually matter.

Whether the reserve increase gets tied to performance vesting. If it does, the plan was always going to be modified and the campaign was theater with a purpose. If it does not, the company genuinely believes it is in a compensation war it cannot win on cash.

Whether a token-side lockup or governance claim appears in the settlement terms. That would be the most informative outcome in the entire affair, because it would mean equity holders have recognized that the real dilution is on the other cap table and have extracted a concession that no securities filing currently requires them to receive.

And whether SkyAI's next capacity agreement is announced before or after the annual meeting. A large, long-dated compute commitment signed while the board is contested would tell you the counterparties are pricing governance risk at zero — which in this cycle has historically been the last comfortable moment.

The plumbing does not show up in the proxy. It never does. But if you want to know who really controls SkyAI twelve months from now, stop reading the compensation schedule and start reading the emission curve. That document has no vote, no dissident, and no fiduciary duty — and it will still be diluting shareholders long after the proxy fight is a footnote.