a16z’s 112-Year-Old Law Problem Threatens Databricks’ $190B Board Seat
A $190 billion Databricks board seat has exposed venture capital’s favourite lie: conflicts are manageable. They are—until your investor knows your rival’s strategy too.
Venture capital’s favourite lie is that conflicts are manageable
A $190 billion Databricks board seat has exposed venture capital’s favourite lie: conflicts are manageable. They are manageable right up until your investor may know what your rival is planning behind closed doors.
The US Department of Justice’s reported investigation into Andreessen Horowitz is a warning shot at the whole AI investment racket: you may be able to back every promising company in a category, but you should not expect to sit in all their boardrooms once they start trying to kill each other.
This is not some theoretical compliance lecture from people who have never built anything. It goes to the core of how modern venture capital works. The biggest firms write cheques across a sector, collect board seats as proof of their value, then tell founders their network is an advantage. Usually, it is. Until the network becomes a pipe carrying sensitive information between rivals.
That is the issue now hovering over a16z. Bloomberg reported that the DOJ has been examining whether the firm’s partners improperly held board roles at competing AI-related data companies. The specific arrangement reported involves a16z co-founder Ben Horowitz on Databricks’ board and a16z partner Martin Casado on Fivetran’s board. ([news.bloomberglaw.com](https://news.bloomberglaw.com/business-and-practice/andreessen-horowitz-focus-of-doj-probe-over-board-directors?utm_source=openai))
The uncomfortable bit is not that investors own stakes in more than one company. That has been normal in venture capital for a long time. The uncomfortable bit is governance. Ownership gives you upside. A board seat can give you the operating picture: where growth is real, where it is fake, what customers are demanding, what product is coming and where management is worried.
Those are two very different levels of access. Founders need to understand the difference before they confuse a prestigious investor name with an aligned director.
Databricks and Fivetran did not begin as the obvious problem
That distinction matters.
Databricks, valued at $190 billion according to reporting on the probe, is best known for helping companies store, organise and analyse large volumes of data. Fivetran built its business moving data between applications and databases. Those sound adjacent, not identical. But markets do not stay politely separated, especially when every company in enterprise software has slapped “AI” on the bonnet and started expanding into its neighbour’s lane. ([techcrunch.com](https://techcrunch.com/2026/08/18/dojs-probe-into-andreessen-horowitz-over-board-seats-baffles-vcs/?utm_source=openai))
Databricks’ Lakeflow product has pushed into AI data pipelines and application connectors — territory close to Fivetran’s core business. The overlap sharpened further after Fivetran combined with dbt Labs in June. Casado had also been on dbt Labs’ board. ([techcrunch.com](https://techcrunch.com/2026/08/18/dojs-probe-into-andreessen-horowitz-over-board-seats-baffles-vcs/?utm_source=openai))
This is how a normal venture portfolio turns into a legal headache. Nobody needs to wake up one morning and decide to fund direct competitors. Startups pivot. Products expand. A feature becomes a product, a product becomes a platform, and suddenly two businesses that once sold different shovels are competing for the same gold field.
That last part is what founders and investors routinely underweight. They assess conflicts based on the company pitch deck from the last financing round. They should assess them based on where the company could reasonably be heading once product, distribution and customer demand start pulling it into the next category.
The reported DOJ scrutiny invokes Section 8 of the Clayton Act, a 1914 antitrust law that restricts interlocking directorates — essentially, the same person serving as a director or officer at competing corporations. It is a 112-year-old rule, and it has rarely been applied to venture capital in this way. ([axios.com](https://www.axios.com/2026/08/18/doj-andreessen-horowitz?utm_source=openai))
Before anyone gets carried away: an investigation is not a finding of wrongdoing. The DOJ has not publicly confirmed the probe, and a16z, Databricks and the DOJ either declined comment or did not respond to the relevant reports. ([axios.com](https://www.axios.com/2026/08/18/doj-andreessen-horowitz?utm_source=openai))
But that does not make this a nothingburger. It makes it more interesting.
The real test is not whether someone can draw a neat line between two companies today. The test is whether the people in each boardroom can still speak freely when those companies begin chasing the same customers, building into the same workflow or competing for the same strategic ground.
The real asset on a board is not advice — it is information
Founders like to hear that their investor will roll up their sleeves, make introductions and help recruit the next executive. Fine. All useful.
But the real value of a board seat is access to the bits nobody puts in the investor update: pricing plans, customer churn, upcoming product moves, hiring problems, acquisition targets, capital needs, the deals going badly and the deals that could change the company.
You do not need a cartoon villain passing a manila folder across a table for this to matter. Information leaks through pattern recognition, casual comments, internal assumptions and the simple fact that smart people cannot unknow what they know.
That is why the phrase “Chinese wall” should not make founders relax automatically. Reporting on the a16z situation noted that the firm could theoretically wall off Horowitz and Casado from each other. Maybe that works. Maybe it does not. But the actual commercial question is brutally simple: if your investor sits on a rival’s board, do you still speak with the same freedom when the room is closed?
If the honest answer is no, then the structure is already hurting the company — regardless of whether a regulator ultimately proves a statutory breach.
That is the bit too many board conversations avoid. A founder does not need to prove that confidential information was shared before raising the issue. If management starts sanitising board materials, withholding early strategy discussions or second-guessing what can be said, the board is no longer doing its job properly.
Venture firms have become comfortable arguing that holding investments in rival startups is normal. In a hot market, it is. Investors want exposure to the winners before anyone knows who they are. TechCrunch noted that multiple venture firms have backed both OpenAI and Anthropic, for example. But owning shares in competing businesses is not the same thing as having formal governance access to both. ([techcrunch.com](https://techcrunch.com/2026/08/18/dojs-probe-into-andreessen-horowitz-over-board-seats-baffles-vcs/?utm_source=openai))
That is the line this investigation could force the industry to take seriously.
It is also a line founders should take seriously without waiting for a regulator to force the point. A board seat is not merely a badge of confidence from a big-name fund. It is a decision about who gets the full version of your business.
The overlooked problem: AI makes portfolio conflicts happen faster
The lazy take is that this is an antitrust headache for a16z. The bigger story is that AI is turning nearly every software category into one giant knife fight.
A data warehouse becomes an AI platform. A data connector becomes an AI pipeline. A coding tool becomes a software development platform. A chatbot becomes a customer-service business. The boundaries move so quickly that an investor’s clean portfolio map can become a mess before the next board meeting.
That puts firms such as a16z in an awkward spot. Their whole pitch is often that they can invest early, invest broadly and help founders navigate a massive network. If regulators begin looking hard at overlapping board seats, the largest firms may become less willing to take seats at all — particularly in sectors where convergence is inevitable.
Axios made the point neatly: if Section 8 is applied more aggressively to venture capital, firms could take fewer board seats, which creates a separate set of concerns. ([axios.com](https://www.axios.com/2026/08/18/doj-andreessen-horowitz?utm_source=openai))
I will go one step further. That could be good for founders who are sick of performative governance, and bad for inexperienced founders who genuinely need an operator-investor in the room.
A board seat is not a trophy. It is a job. The right investor can stop you making a ruinous hire, save you from a fantasy valuation, force you to face a broken sales model, or make the introduction that changes the business. The wrong investor is a spectator with privileged access.
If big venture firms retreat from boards to avoid conflicts, founders will need to become much more deliberate about who fills that gap. They may need independent directors earlier. They may need former operators instead of logo-heavy investors. And they will need better governance before the business is big enough to hire expensive adults to clean up the mess.
This is the practical trade-off. Less board involvement from major funds may reduce one kind of conflict, but it does not automatically give founders a better board. A weak investor seat replaced by nobody is not governance. It is just an empty chair while the founder makes harder decisions alone.
The answer is not to treat every portfolio overlap as a scandal. The answer is to recognise that AI makes overlap predictable, then build a process for dealing with it before trust gets damaged.
The contrarian view: this might improve venture capital
Most VCs seem baffled that the DOJ would focus on this. I understand why. There are plenty of bigger, uglier fights in AI: concentration of compute, data-centre financing, cloud power, government contracts and platform lock-in.
But small governance rules shape big markets.
If a handful of elite funds can hold board influence across businesses that drift into direct competition, then founders may get capital while quietly surrendering strategic independence. That is not always malicious. It is often just the predictable result of capital trying to maximise optionality.
The best outcome here is not that regulators make venture investing impossible. That would be idiotic. The best outcome is clarity: when a portfolio overlap becomes material, someone steps down, the company gets an independent director, and the rules are decided before confidential strategy becomes a grey-area problem.
That clarity needs to be commercial as well as legal. Who receives the board pack? Who joins closed sessions? Who is excluded when a particular competitor, acquisition target or product move is discussed? What happens if two portfolio companies move closer together after the investment has already been made?
Those are not glamorous questions. They are adult questions. And they are far cheaper to answer while everyone is getting along than after a founder believes their investor has divided loyalties.
That is not anti-venture. It is pro-founder and pro-competition.
And frankly, it is better business. A founder should know exactly who hears what, who owes duties to whom, and whether their investor can still advocate for them without divided loyalties. If that question makes people uncomfortable, good. It should.
What this means for you
If you are a founder, do not wait until your company is worth billions to treat board design seriously.
First, ask every current and prospective investor a direct question: Which companies in your portfolio could become competitors in the next 24 months? Not competitors today. Competitors after the obvious product expansion everyone is pretending not to see.
Then ask the harder follow-up: If that happens, what changes? Do not accept vague reassurance about relationships, internal process or good intentions. Find out whether the investor will keep the seat, step back from particular discussions, or accept that someone may need to leave the board.
Second, get specific about information rights. Your board pack should not become a broadly circulated newsletter for every investor with a fancy title. Decide what is genuinely board-level, what is observer-level and what should stay with management until it is ready.
Third, add at least one director whose incentive is simple: make your business better. Not their fund’s portfolio. Not the next financing round. Yours.
Finally, review the portfolio-conflict question whenever your product changes materially, not only when you raise money. The company you are competing with next year may not be the company anyone named in the last board discussion.
And if you are an investor, stop selling board seats like they are frequent-flyer upgrades. Take the seat only if you will do the work, can remain fully loyal when markets converge, and are prepared to leave when you cannot.
That is the adult version of venture capital. The DOJ’s reported a16z probe may not end in fireworks. But it has already exposed the question every founder should ask before signing the paperwork: when my investor owns a piece of my rival, whose side are they really on?