What control rights can the founding team not afford to lose when an AI startup raises funding?

According to reports from media outlets such as The Wall Street Journal, DeepSeek recently completed its first major round of financing, with a scale of approximately $7.4 billion and a post-investment valuation exceeding $50 billion. The reports also mention that Liang Wenfeng personally invested about $3 billion, and most external investors did not directly invest in DeepSeek's operating company but entered through a limited partnership managed by the founder, subject to arrangements such as a five-year lock-up period.

It should be noted that the above information primarily comes from media reports, and DeepSeek has not yet publicly disclosed the complete transaction documents. Therefore, this article does not comment on the authenticity of the transaction nor attempt to reconstruct DeepSeek's full governance structure. Instead, based on these media reports, it discusses a more general question: How can AI founders retain key decision-making capabilities, provide reasonable protection to investors, and avoid turning the company's operational risks into personal risks when introducing substantial capital?

For founding teams, the truly dangerous question during fundraising is often not "how much am I being diluted," but rather: after the next round of funding comes in, who decides whether the company sells, whether the technology roadmap shifts, whether the CEO is replaced, whether core assets can be licensed out, and who bears the repurchase liability if the project fails.

Why AI Companies Value Control More Than Ordinary Startups

AI companies typically require sustained large capital investments. Model training, inference computing power, core talent, data acquisition, and product commercialization can all incur long-term expenses. Relying solely on the founding team's own funds often makes it difficult to complete the entire process from R&D to scaled application.

At the same time, the technology roadmap and business cycle of AI companies carry significant uncertainty. Investors are usually more concerned about revenue, valuation, and exit cycles, while founding teams may prioritize model capability, open-source strategy, and long-term R&D. The two sides are not inherently in conflict, but at critical junctures, their judgments may differ markedly.

The transaction window in the AI industry may also be shorter than in traditional sectors. After a product reaches a certain stage, it may need to be sold promptly to a large tech company, or engage in M&A or deep collaboration with industry giants; alternatively, it may require abandoning the original direction and quickly pivoting to a new model, scenario, or business model.

These major decisions often cannot be debated indefinitely. If the governance structure is too fragmented, each decision—such as a pivot—requires repeated negotiations among multiple parties, and the company may miss the optimal window.

Therefore, retaining decision-making authority over key matters for those who truly understand the technology, product, and team helps the company make efficient decisions at critical moments. But efficient decision-making does not mean founders are unconstrained. The value of control is to enable the company to make clear decisions, not to allow founders to ignore the interests of the company, investors, and other shareholders.

What the Limited Partnership Solves

If investors invest directly in a startup, they typically become shareholders in the company and may demand board seats, among other things.

A limited partnership offers an alternative arrangement: investors first put their funds into a holding vehicle, which then invests in the operating company. At the shareholder level of the operating company, multiple originally dispersed investors may be consolidated into a single holding entity.

Investors mainly obtain economic benefits, information rights, oversight, and specific protections within the limited partnership (these information rights and corresponding protections require reasonable agreements balancing the interests of all parties); how the limited partnership exercises shareholder rights at the operating company level is governed by the partnership agreement.

If the DeepSeek transaction structure disclosed by the media is accurate, this arrangement may reduce the degree of direct involvement by external investors in day-to-day operations and technology roadmaps, while enabling the founder to coordinate decision-making within the holding vehicle.

But a limited partnership is not a magic bullet for control.

Controlling a holding vehicle only means the founder can potentially influence how one shareholder exercises its rights; it does not necessarily mean they control the entire company. True corporate control also depends on who decides key matters, who appoints and removes the core team, and whether these arrangements can be sustained in subsequent financing rounds.

Control Is Not Just About Equity

Founding teams can assess whether they truly have control by looking at three questions.

The first question is: who ultimately decides matters that change the company's fate?

For example, whether the company continues fundraising, sells or merges, transfers core models, code, data, and intellectual property, changes its technology roadmap, or discontinues existing business. Even if the founding team still holds a high percentage of equity, if these matters are subject to unilateral veto by investors, their actual decision-making space may still be significantly limited.

The second question is: who decides the core team?

Who can appoint or replace the CEO, CTO, and key management personnel? Who determines R&D budgets and commercialization pace? These often affect the company's actual direction more than equity percentages themselves.

The third question is: can the current control arrangements be sustained?

Having decision-making power today does not mean it will remain after the next financing round. New investors' veto rights, changes in board seats, continuous equity dilution, and mechanisms for replacing the managing partner (the one with decision-making authority) within the holding vehicle can all alter the control status.

Therefore, the founding team needs to design not an isolated holding vehicle but a complete decision-making chain: through which entity funds enter, how that entity exercises shareholder rights, who decides major matters, who appoints and removes the core team, and whether these arrangements remain valid after subsequent financing rounds.

Which Decision-Making Powers Should Not Be Easily Ceded?

Founding teams do not need absolute decision-making power over everything. Investors having visibility into the company's financials, monitoring the use of investment funds, and restricting improper related-party transactions are generally reasonable protections.

What truly requires careful ceding is decision-making power over matters that could change the company's fate.

First, the disposition of core technology and intellectual property. Models, code, training outputs, data resources, and core patents are typically the most important assets of an AI company. For the sale, exclusive licensing, or transfer of these assets, the founding team should at least retain joint decision-making power or a necessary veto.

Second, company sales, mergers, and major transformations.

When an AI product reaches a certain stage, selling to a large company does not necessarily mean startup failure; it can also be a reasonable path for the product, team, and technology to continue developing. But when to sell, whether to sell equity or technology, and whether the founding team stays on—these need input from those who truly understand the company's long-term value.

Third, the appointment and removal of the core management team. If investors can easily replace the founder, CEO, or core technology lead under normal circumstances, the founder's control may become merely formal.

Fourth, subsequent financing and significant dilution. A new financing round not only changes equity percentages but may also reallocate board seats, veto rights, and management appointment powers. The founding team should not only calculate equity percentages after this round but also simulate the governance state after the next round or even the one after that.

Founding teams should not broadly pursue "I decide everything." A more realistic approach is to first list three to five core decision-making powers that cannot be easily lost, and then grant investors reasonable protections on matters like financial oversight, information disclosure, and risk control.

Retaining Control Does Not Mean Excluding Investor Protection

When investors are willing to invest without directly controlling day-to-day operations, they typically demand protections at other levels, such as understanding operational and financial conditions, monitoring the use of investment funds, limiting self-dealing, obtaining remedies in case of serious breaches, and having exit options upon reasonable conditions or milestones.

Founders should distinguish between two types of arrangements.

One type is control over product direction, technology roadmap, and daily operations; the other is protective rights to prevent asset transfers, conflicts of interest, and serious breaches.

The former can be more retained within the founding team; the latter should leave reasonable space for investors. A mature financing structure is not one where investors can only provide funds and watch, but one that reduces their direct interference in business direction while safeguarding their basic rights to information, oversight, exit, and remedies.

In some valuation adjustment mechanism disputes I have handled, I have seen many cases where the founding team did not grant investors necessary information and oversight rights, nor reserved reasonable exit mechanisms, yet signed even stricter repurchase, guarantee, pledge, and default liability clauses to compensate. On the surface, they retained operational control, but in reality, they may have turned the company's operational risks into the founder's personal debt risks.

The more concentrated the control, the more the founding team needs to emphasize corporate independence. Registering core intellectual property in individual or affiliated company names, mingling personnel, data, and R&D outputs across multiple entities, and transferring funds and business opportunities through related-party transactions can all turn control issues into corporate governance and personal liability problems.

Before Signing a Term Sheet, Founders Should at Least Think Through Five Questions

Valuation and dilution percentages are important issues, but financing negotiations should not revolve solely around these. Before signing a term sheet, the founding team should at least think through the following five categories of issues.

First, control

Which matters require the founding team's consent? Which matters can be decided by a majority of the board? Which matters can be delegated to the management team? In particular, list separately: company sale, major financing, disposal of core assets, changes in technology roadmap, and appointment/removal of core management.

Second, investor protection

Which information rights, oversight rights, and exit rights can be granted to investors? Which veto rights might materially affect day-to-day operations? Do not just look at the clause names; examine the trigger conditions, scope of application, and the actual effect of overlapping rights.

Third, repurchase clauses

Will you accept repurchase clauses? Which events may trigger repurchase obligations of the company, founder, or controlling shareholder? Failure to go public, unmet performance targets, failure of subsequent financing, compliance issues, and founder departure—do all of these lead to repurchase? Is the repurchase price compounded with fixed returns, compound interest, and high penalty fees?

Fourth, the holding vehicle

Who serves as the managing partner? Who can replace that entity? What oversight, exit, and remedy rights do limited partners have? The holding vehicle should not only facilitate entry during this financing round but also allow smooth adjustments in subsequent financing, M&A, IPO, and exit scenarios.

Fifth, dispute resolution

This is a very core issue in practice but often overlooked in the initial stages. Do the investment agreement, shareholders' agreement, repurchase agreement, guarantee contract, and holding vehicle agreement have interconnected dispute resolution arrangements? When multiple entities are involved, is there a procedural split where some disputes go to litigation and others to arbitration?

The significance of these five questions is to translate the abstract concepts of "founding team control" and "investor protection" into concrete, negotiable transaction terms.

What May Be More Dangerous in Financing Terms Is Not Equity Percentage

In the dozens of equity investment, repurchase, and commercial arbitration cases I have been involved in and handled, many founding teams focused most on valuation, amount received, and dilution percentage during financing, but underestimated the long-term risks of repurchase, guarantee, and dispute resolution clauses.

Repurchase clauses require particular caution.

Founding teams should confirm: does the repurchase obligation fall on the company, the founding team, or the controlling shareholder? Is there joint liability? Are the repurchase trigger conditions too broad? Does the repurchase price include fixed returns, compound interest, and high penalty fees? Does the founder have realistic ability to perform?

Once the founder personally assumes a repurchase obligation, the risk of the company's project failure can directly transform into personal asset risk. When the company bears the repurchase obligation, one cannot just look at the contract terms; it must also be assessed under corporate capital maintenance rules and specific performance conditions.

Dispute resolution clauses should not be handled lightly either.

Equity investment, control, and repurchase disputes often involve trade secrets, multiple transaction documents, and multiple related entities. Commercial arbitration, with features like non-public hearings and relatively flexible procedures, is typically more suitable for complex commercial disputes.

However, choosing arbitration also requires caution. The arbitration institution, seat, applicable rules, language, and scope of disputes should all be carefully evaluated before selection. Otherwise, the founding team or investors may face a difficulty and cost of enforcement they did not truly understand when signing the agreement. For example, in some blockchain-related investment disputes I have handled, the cost of enforcing rights during the arbitration phase exceeded what the rights holder could accept, thus significantly reducing the possibility of the breaching party bearing its liability.

Conclusion

For AI founders, financing documents are not a set of legal templates but an operating system that determines the company's decision-making power, risk allocation, and exit paths for years to come.

A truly mature financing structure is not one where founders are never constrained, nor where investors can only watch from the sidelines. Rather, it clarifies in advance: which matters are decided by whom, which risks are borne by whom, which events trigger exit, and which mechanism resolves disputes if they arise.

Before capital enters the company, control, repurchase obligations, holding vehicle, and dispute resolution mechanisms should be designed simultaneously. Otherwise, the day the financing closes may also be the day future control disputes and personal liability risks begin to accumulate.

Original Author: Zhao Xuan