According to reports by media outlets including The Wall Street Journal, DeepSeek recently completed its first major round of financing, raising approximately $7.4 billion, with a post-money valuation exceeding $50 billion. Related reports also note that Liang Wenfeng personally invested about $3 billion. Most external investors did not invest directly into DeepSeek’s operating company; instead, they entered via a limited partnership managed by the founders, subject to arrangements such as a five-year lock-up period.
It should be noted that the above information currently stems primarily from media reports—DeepSeek has not yet publicly released complete transaction documents. Therefore, this article does not assess the authenticity of the transaction nor attempt to reconstruct DeepSeek’s full governance structure. Rather, it uses this media-reported case to discuss a broader question: When AI founders raise massive capital, how can they retain critical decision-making authority, provide reasonable protections to investors, and avoid transforming the company’s operational risks into personal liability?
For founding teams, the truly dangerous questions during fundraising are often not “How much have I been diluted?” but rather: After the next round of funding, who decides whether the company is sold, whether the technical roadmap shifts, whether the CEO is replaced, whether core assets can be licensed out—and who bears the buyback obligation if the project fails?
AI companies typically require sustained, substantial capital investment. Model training and inference compute, top-tier talent, data acquisition, and product commercialization all entail long-term expenditures. Relying solely on the founders’ own capital often proves insufficient to carry the company through the entire journey—from R&D to scaled application. At the same time, AI firms face considerable uncertainty in both technical direction and commercial cycles. Investors usually prioritize revenue, valuation, and exit timelines, while founding teams may place greater emphasis on model capability, open-source strategy, and long-term R&D. These perspectives are not inherently conflicting—but at critical junctures, judgments may diverge significantly.
The deal window in AI may also be shorter than in traditional industries. Once a product reaches a certain stage, it may need to be promptly sold to a major tech firm, merged with or deeply partnered with an industry giant—or, conversely, it may require abandoning the current path and rapidly pivoting to a new model, use case, or business model. Such pivotal decisions often cannot be deferred indefinitely. If governance structures are overly fragmented, every strategic pivot or major decision would require repeated negotiations among multiple stakeholders—risking the loss of optimal timing. Thus, retaining decision-making authority over key matters with those who genuinely understand the technology, product, and team helps ensure efficient, timely decisions at critical moments. Yet efficiency does not equate to unconstrained founder control. The value of control lies in enabling clear corporate decisions—not in allowing founders to disregard the interests of the company, investors, and other shareholders.
When investors directly fund a startup, they typically become shareholders outright and may demand board seats, among other rights. A limited partnership offers an alternative arrangement: investors first commit capital to a holding vehicle (the LP), which then invests in the operating company. At the shareholder level of the operating company, numerous otherwise-dispersed investors may thus be consolidated into a single legal entity. Investors derive economic returns, oversight rights, and specific safeguards primarily within the LP structure; how the LP exercises its shareholder rights in the operating company is governed by the partnership agreement. If the DeepSeek transaction structure reported by the media is accurate, this arrangement may reduce the extent to which external investors directly intervene in day-to-day operations and technical strategy—while enabling founders to centrally coordinate decisions across the holding vehicle. However, a limited partnership is no “control magic.” Controlling a holding vehicle only means founders may influence how one shareholder exercises its rights—not that they automatically control the entire company. True corporate control depends on who decides key matters, who appoints and removes the core management team, and whether those arrangements remain enforceable in subsequent funding rounds.
Founding teams assessing whether they truly hold control should begin by asking three questions: First, who ultimately decides matters that change the company’s fate? Second, who appoints and removes the core team? Third, can the current control arrangements endure? Accordingly, founding teams must design not an isolated holding vehicle—but a complete decision chain: through what entity does capital enter; how does that entity exercise its shareholder rights; who decides major matters; who appoints and removes the core team; and whether these arrangements remain effective after the next round of financing.
Founding teams need not retain absolute decision-making authority over all matters. Investor rights to review financials, monitor fund usage, and restrict unreasonable related-party transactions are generally reasonable protections. What truly warrants careful reservation, however, are decision rights that could alter the company’s destiny. These include, first, disposition of core technology and intellectual property; second, company sale, merger, or major strategic pivots; third, appointment and removal of the core management team; and fourth, follow-on financing and material dilution. Founders should avoid the blanket goal of “I decide everything.” A more realistic approach is to first identify three to five non-negotiable core decision rights—then grant investors reasonable safeguards on financial oversight, disclosure, and risk mitigation.
Retaining control does not mean excluding investor protections. Investors willing to commit capital without direct operational control typically seek protections elsewhere—for example, access to operational and financial information, monitoring of fund usage, restrictions on self-dealing, remedies upon material breach, and exit options upon achieving reasonable timeframes or conditions. Founders should distinguish between two categories of provisions: one governing product direction, technical roadmap, and daily operations (control rights); the other preventing asset transfers, conflicts of interest, and material breaches (protective rights). The former can reasonably remain with the founding team; the latter should afford investors appropriate scope. A mature financing structure does not relegate investors to passive capital providers—it reduces their direct interference in operational direction while safeguarding their fundamental rights to information, oversight, exit, and redress.
Before signing a term sheet, founding teams should at minimum clarify five categories of issues: control rights, investor protections, redemption clauses, holding vehicles, and dispute resolution. These five points translate the abstract concepts of “founder control” and “investor protection” into concrete, negotiable transaction terms.
What poses greater danger in financing terms may not be equity percentage alone. In dozens of equity investment, redemption, and commercial arbitration cases we’ve advised on or handled, many founding teams focus intensely on valuation, disbursement amount, and dilution ratio—yet underestimate the long-term risks embedded in redemption, guarantee, and dispute resolution clauses. Redemption clauses, in particular, warrant especially careful scrutiny. Dispute resolution clauses likewise must never be treated casually. Equity investment, control, and redemption disputes often involve trade secrets, multiple interrelated agreements, and numerous affiliated entities. Commercial arbitration—characterized by confidential hearings and relatively flexible procedures—is typically better suited for resolving complex commercial disputes.
For AI founders, financing documents are not merely boilerplate legal templates—they constitute an operating system that determines corporate decision-making authority, risk allocation, and exit pathways for years to come. A truly mature financing architecture neither leaves founders perpetually unaccountable nor confines investors to passive observation. Instead, it proactively clarifies: who decides what, who bears which risks, what triggers an exit, and which mechanism resolves disputes when they arise. Control rights, redemption obligations, holding vehicles, and dispute resolution mechanisms must all be designed in tandem, before capital enters the company. Otherwise, the day the financing closes may also mark the beginning of accumulating future control disputes—and personal liability exposure.
[Zhao Xuan]
When AI startups raise massive capital, how can they retain critical decision-making authority, provide reasonable protections to investors, and avoid transforming the company’s operational risks into personal liability? The answer lies in the intricate balance between control rights, investor protections, redemption clauses, holding vehicles, and dispute resolution mechanisms.
The DeepSeek deal structure, as reported by media outlets, offers a glimpse into the complex dynamics at play. By retaining control over key decision-making processes through a limited partnership arrangement, the founding team can balance the interests of investors with the need to make strategic decisions that drive growth. This structure allows the founders to exercise control over the company’s direction while also ceding reasonable safeguards to investors.
However, it’s crucial to note that the true value of control lies not in exercising unconstrained power but in enabling clear corporate decisions that benefit the company, its investors, and other shareholders. Founding teams must design a complete decision chain that clarifies who decides key matters, who appoints and removes the core management team, and whether these arrangements remain effective after subsequent rounds of financing.
Retaining control does not mean excluding investor protections; rather, it requires fostering a mature financing architecture that reduces direct interference in operational direction while safeguarding investors’ fundamental rights to information, oversight, exit, and redress. This calls for careful consideration of five key issues: control rights, investor protections, redemption clauses, holding vehicles, and dispute resolution.
Equity percentage alone may not pose the greatest danger in financing terms. Founding teams should prioritize understanding the long-term risks embedded in redemption, guarantee, and dispute resolution clauses. Redemption clauses, in particular, warrant careful scrutiny, while dispute resolution clauses should never be treated casually.
Commercial arbitration cases and other disputes often involve trade secrets, multiple interrelated agreements, and numerous affiliated entities, making it essential to design financing documents as an operating system that determines corporate decision-making authority, risk allocation, and exit pathways for years to come.
Ultimately, the goal is to proactively clarify who decides what, who bears which risks, what triggers an exit, and which mechanism resolves disputes when they arise. Control rights, redemption obligations, holding vehicles, and dispute resolution mechanisms must all be designed in tandem, before capital enters the company. Only by doing so can AI founders truly leverage their control rights to drive long-term growth and success.