The round on the cover

I was talking about recent seed rounds with a friend the other day, discussing how they seemed to be inflating in size, with many creeping up to $10M of initial capital in. I’ll save any judgment, but this is an increasing departure from traditional seed round sizes. My sense is that most founders don’t target these higher amounts from the jump — they still set out to raise $2M - $3M seeds, but there’s now a wider variance in outcomes of where those rounds land. Internally, we call this “the round on the cover,” which can end up differing from the final amount that appears in TechCrunch, often bubbling up far beyond the initial proposal. 

If you talk to people about it, they’ll just tell you the dynamic is crazy — emblematic of frothy market conditions. Some might say it’s just another reshuffling of round names, in the way that pre-seed was the new seed, and seed was the new Series A a few years ago. I thought I’d peel back at the incentives on both sides of these transactions — founders and investors — to land at an understanding of what’s going on. 

I don’t know how many founders plan on their round sizes expanding way beyond their starting threshold, but for founders of a certain pedigree operating in defined categories, a lower number on the cover can work to their advantage. It reduces the allocation available for investors, thereby increasing fomo in an already capital-pervasive environment. In some instances, the founders will close up a round, but (sometimes uncapped) notes or SAFEs get stacked on top. This is a tricky dynamic and a separate discussion altogether. 

In some cases, traditional seed funds catalyze larger raises; but in many instances, a larger fund or capital base will lead these bigger seeds. Larger venture firms started doing seeds as they felt pressure from crossover funds and late-stage activity to move earlier in the fundraising lifecycle. The bind lies in the fact that many conventional Series A or B shops actually have even larger funds to deploy with distributions from the past decade coming back, but their investors are now allocating more time to places where they had less money at work — maybe $1M - $3M investments out of $500M+ vehicles. Furthermore, ownership thresholds at the early seeds are coming down, and if the company shows any sign of promise, there’s no guarantee that the big fund that leads the seed can also lead the A (to build more ownership in the company over time). There also might be a newer implicit assumption about how risk curves work for startups, asserting that steps between the seed and the A don’t change the risk profile of the company on a consequential level. I recommend Kanyi’s piece on this if it’s a topic of interest. He has the crispest articulation I’ve seen. 

This all leads me to pose the (perhaps unfairly simplistic) question: would you rather invest $2M into a startup for 10% ownership or invest $10M to guarantee 20% ownership, putting more dollars at work out of a fund with limited to no additional risk? There are a lot of assumptions baked into this prompt, but it demonstrates the rough calculus from the investor perspective and the incentive to offer a larger amount to the founder.

Again, I’ll spare the judgement on whether this is right or wrong — it’s ultimately a market, a scaffolding for incentives. It feels a little bit like I’m giving away a trade secret, but I continue to reflect on what I see in the current market and the moves that happen with a subset of seed raises. 

Areas of Interest, March 2021

I used to write more posts or essays covering ideas, trends, and themes that were compelling to me in the moment. In the current market, there may be an abundance of capital, but I’ve also found a plethora of startups and entrepreneurs excavating in exciting places I would never have considered. This has pushed me to be more reactive to the broader ecosystem, setting aside the usual presumptions. Nevertheless, I’ve always enjoyed taking a step back to publicly share what I’m thinking about to see if my interests (usually quite obscure) are aligned with anyone else’s.

The interplay between code and visual workflows

Others have written about this, but more enterprise processes and workflows are increasingly defined in code, turning configuration files and scripts into templatized, repeatable logic. Examples include cloud native security (Open Policy Agent), business operations (Salto), and infrastructure provisioning (Terraform). By a different token, I’ve also seen new projects that help users define, create, or interact with various processes through visual builders. Terrastruct in software architecture or Tonkean in process automation come to mind. The historical knock against visual workflow tools (WYSIWYG website builders like Squarespace come to mind) is that they wouldn’t be able to sufficiently capture the complexity of the system in question, especially the edge cases. Now, “code” can act as an escape hatch when a UI isn’t enough. Design tool Plasmic is an example. State machines are also an avenue to explore. My underlying thesis is that many parts of the enterprise (from website to API design) can be produced through the combination of visualizations and code, not isolated to either or.

Industrial biotechnology

What would Dow or DuPont look like if they were started today? Multinational chemical companies like these produce products that touch many parts of the economy, from agriculture to materials to therapeutics. Yet, the chemical industry is one of the most energy intense in the world with significant carbon contributions. Alternative solutions with the potential to be cheaper, safer, and more sustainable have begun to emerge. They’re typically defined by biosynthesis, where enzymes catalyze reactions to create compounds, standing in opposition to the legacy chemical processes. Enzymes have been part of industrial processes in various forms for some time, but companies like Solugen now use computational protein design methods to identify specific catalysts. Pow is another interesting platform in this space. Traditional software venture investors dipping their toes into biotech can be a slippery slope, but the expansive possibilities that come as a result of decreasing cost curves, along with the decarbonization of a global engine industry, excite me.

Machine learning systems as an enterprise attack vector

In the enterprise, new technologies bring forward the promise of business acceleration, but they also introduce the possibility of new attack vectors. In turn, this domino effect contributes to the emergence of new security businesses. This was evident in cloud security (Lacework, Orca) or IoT (Armis) a couple years ago, and now has taken hold in the API space (Salt, Noname). I’m the first to admit “machine learning” is a loaded term, and there exists some debate about the level to which ML platforms have penetrated into enterprise accounts. However, over the next decade, ML systems will increasingly go into production as the largest banks & financial institutions will use it to fight fraud, retailers & grocers will deploy personalized recommendations for their customers, industrial companies & manufacturers will optimize supply chains...the list could go on. These activities only increase the surface area for misuse and cyberattacks, with one example being adversarial machine learning. The risk is so pronounced that leading research institutions and tech companies co-published the Adversarial ML Threat Matrix. Robust Intelligence is a startup working on defending operational AI systems. 

If you’re interested in exploring any of these topics or working on something in one of these categories, I’d love to chat. Email me at aashay [at]

The nature of "the miss"

In December, I spent time thinking about the investing work I had done coming to that point. I gleaned lessons from companies we were lucky to be involved with where something was working, but candidly, I beat myself up over perceived misses — companies I had seen at pre-seed or seed that weren’t in the portfolio who had some early signs of success. The rationale for not investing varied, spanning a lapse in judgment to a lack of experience in the category. These investments bit at me not only because the strength of the founders started to shine through over a longer period of time, but also because the full potential of the market became more legible in hindsight. It absolutely killed me.

So much of the venture business is guided by opportunity cost (in many ways, it defines it), and I was starting to internalize that. I’m not as concerned about errors of commission - I’d like to push to a place on the risk curve that feels apt for the asset class and support companies that reach global, not local, maxima.

But, focusing too much on supposed mistakes and over-extrapolating can be dangerous. I’ve come around on my thinking in recent weeks and am approaching “misses” with a different lens. 

  1. Opportunity cost is contextual and based on firm dynamics. As fund AUM grows, the paths to returning it shrink narrowly with increased pressure. Missing Uber or Airbnb hurts more and more. Specialization can also grow the burden of proof, especially if winners stand out in certain categories. I’m lucky to be a generalist early stage investor, which alleviates some of these concerns. 

  2. There is a high degree of markup activity in the ecosystem, but there’s little to point to on the durability of the outcomes. While Series A and Series B raises do de-risk companies, the end results may be years away. And that’s the ultimate concern, not TVPI. Trust the inputs.

  3. Over-indexing on missed investment opportunities could contribute to even more FOMO-driven investing. Part of venture does rely on this; it’s quite difficult to be 100% contrarian and succeed in VC, but I know that I personally do not want to be heat-chasing all the time. Investing edge has three components — informational, analytical, and behavioral. If the first is closer and closer to dissipating and the second isn’t as applicable to seed, perhaps the third drives alpha. Just a thought.

None of this is to say that misses don’t hurt, but there are likely healthier approaches and framing. For example, I really enjoyed this video with Felicis Ventures’ Aydin Senkut. While he notes that he missed Uber and Airbnb, their success keyed him in on the rise of global payments as a second-order effect and his subsequent investment in Adyen. What can I learn from a miss about what’s working in the world? What are the complements of the opportunity I wish I was involved with?

Borrowing Scale

The adage "software startups are cheaper than ever to start, but as expensive as ever to scale" is firmly embedding itself as part of canonical tech wisdom, as what's accepted to be true. I even tweeted about it myself the other day. APIs like AWS that offset initial capex & higher-level programming languages make getting up and running over a weekend easier and cheaper for a small group of hackers. On the other hand, high-priced engineering talent, operational costs, and distribution efforts drive costs up as the product and company scale.

The question I'd ask — will the latter half of this statement continue to be true in 5 years? 10 years? 20? What could lead to a substantial cost reduction in scaling and growing an organization?

Over the past few years, more vendors are emerging as default options for a company when it raises a seed or A that help with the Day 2 problems. Skip the roundabout with payroll providers and banks — just use Deel to build your overseas team. Keep finance and accounting in check from the start with Ramp. Need to sell into enterprise? Streamline a SOC-2 approval with Vanta.

Before, manual effort and labor fixed and massaged these problems, contributing to the costly business of scaling. Now, startups can effectively can not only borrow scale for launching a product, but also building their organizations. These tools won't replace the staff needed to manage these activities altogether, but they do aspire to give employees more leverage.

Furthermore, startups are also building abstractions on the complex, distributed systems required for software to serve millions of users. Okteto provides developers a slick interface on top of Kubernetes, a lower-level primitive for orchestrating applications. Fauna seeks to make a distributed & reliable operational database available through a single API. All of this is an effort to accelerate development velocity and timelines. While this trope probably won't realize for a long time, tailwinds seem to be pushing us towards Chris Dixon's "unicorn of one."

Some additional thoughts:

  1. What are the second order effects of the cost of scale coming dramatically down? An entire investment and employee ecosystem has been built on this premise alone.

  2. I sense security is an opportunity area to add to this list.

  3. Distribution is one area where costs do not seem to be coming down as channels and buyers are inundated. If this is the primary cost driver, perhaps this whole argument is moot.

Looking Back, Looking Ahead II

Almost exactly a year ago, I wrote a piece titled “Looking Back, Looking Ahead” sharing my reflections after spending ~6 months as an early stage investor. While there was positive reception, the post served as a guidepost for me to orient around and build on. This year, I won’t comment on the macro narratives of 2020, but rather, share more “notes from the road” that have come to light through the past year. The surface area here is limited to personal observations about what I know and learn about — the venture industry, early stage startup ecosystem, and software markets. Each of these vignettes could command an individual post of their own, so think of them more as jumping off points for conversation.

If you’re interested in discussing any of the below, please reach out at aashay[at] 

Venture & Investing

Investment sizing is a function of conviction. Far from an original thought, a close friend shared this with me as I was thinking through different allocation amounts in a round. If you’re wrong about an investment in venture, you lose your principal, but the position size determines the magnitude of losses. Vice versa for gains. While this quip may seem intuitive on its head (partially derived from the Kelly criterion), it has been harder to put into practice because of the questions surrounding conviction. Where do you choose to find it as an early stage investor (given limited data)? How does one define deep vs. narrow conviction? When should one “borrow” vs. form conviction independently? 

Increased investor specialization helps drive valuations. This argument may be overly simplistic, but I’ve seen game theoretic dynamics play out in certain categories (fintech, SaaS, infrastructure) partially driving up early stage valuations. If you view the venture industry through the barbell lens, platform funds are growing larger and larger on one end of the spectrum. A division of labor is the efficient way to organize a large organization, thus the increased presence of specialist investors within big firms. As one constrains the investable universe, it may be easier to rationalize the inevitability of certain outcomes within a category. With that in place, the associated opportunity cost (and career risk) of *not* doing certain deals is too high — you’re left with high prices and auction dynamics. 

Seed is the competitive frontier. Now what? Over the past year, I’ve spoken with a handful of GPs who used to write $8M - $10M checks at the A now comfortable leading seed rounds with $2M - $4M investments to get their foot in the door of a promising company. Traditionally, Series As were the competitive rounds for companies who had recently hit PMF, but that game has leaked into seed. On one hand, this could just be a reshuffling of round names; on the other hand, companies could be capitalizing and structuring ahead of traction in a world where constraints matter.

Early stage venture shares little with other forms of capital allocation (private equity, hedge funds). While early stage investing generally needs a shrewder lens and sense of analytical rigor, newer approaches (in the vein of public company analyses) may also be missing the forest for the trees. Startups are inherently imperfect — one could argue they’re collections of individuals building products, not companies. Investing in them entails finding the right sources of information, unique access, along with crafting & understanding narratives.

How does liquidity follow paper gains? I’m not exactly sure how I would frame or parameterize this research, but one piece of data I’ve wanted access to is the resulting liquidity and cash returns of companies that generated 10x paper markups within certain funds or vintages. Right now, with preemptive rounds and an unwavering belief in the expansion of software markets, markups (and TVPI) may be abundant for some managers, but they’re simply proxies for ultimate success. In other words, intermediate metrics are tough

Personal brand may continue to matter even more for investors, but I personally struggle with the implications of this. Founders now can pick individual investors over institutions, and one way to bypass the traditional gatekeepers to stand out is through a recognizable, unique, and authentic social presence. Some may dismiss this as a flash in the pan, but I’ve seen countless examples of people building real investment careers starting on the Internet, and the trend will continue to grow. My own life wouldn’t look the same if I hadn’t met people on Twitter or written publicly. Yet, I can’t help feeling as if platforms like Twitter have become increasingly noisy and performative (maybe they’ve always been and I’ve just missed something). There’s one question of continuing to stand out in that environment, but if the north stars are investing in and building great companies, what are the best avenues going forward? 

Some investors are taking a “value chain” approach to portfolio construction. I see the benefits. In the 1990’s, when John Doerr was at the peak of his powers, he relied on his form of a California Keiretsu. Originally a Japanese concept that dominated their late 20th century manufacturing prowess, keiretsu means a set of companies with interlocking business relationships. Now, as different “stacks” (data infrastructure, JAM, e-commerce) become more legible, investors complement an initial investment in a category with other members in the value chain. The companies may see some product or partnership benefit from working together, but they can also collectively reinforce a narrative about their corner of the market, thus bringing their future cost of capital down in tandem. 

I admire investors who have nailed their personal process because it’s easy to give lip service to, but incredibly hard to implement. In an ecosystem with a high volume of deal activity, one can get lost chasing shiny objects. The investors I’ve come to admire the most are intensely clear about what they look for an investment, and this clarity compounds on itself through the sourcing, selection, winning, and management phases. For example, read this doc from Brad Gillspie, of IA Ventures. I read and reference it often because it not only has informed my thinking on the mechanics of seed stage startups (bearing a hypothesis-driven mindset, staying capital efficient, “earning your burn”), but also helped structure my approach and personal process. Finding conviction in a process is difficult as a younger investor because you don’t have enough data or historical runway to train your algorithm. As I look forward, finding the path to trust my judgment and making my guiding principles more concrete is high on the priority list. 

Markets & Sectors

What’s being underwritten in fintech infrastructure? Payment processing. Card issuing. Identity verification. Payroll APIs. Fraud detection. Tapping into ACH. Kicked off the Plaid acquisition in January, this has been a banner year for fintech infrastructure investments, especially in the venture market. I’ve asked myself what founders and investors are projecting into the future if the number of neobanks and PFMs that use these services might seem small on its head. Growth is part of the equation, along with digital financial services leaking into other categories (online marketplaces, healthcare), and a belief in the power of accumulated data. I’ve started to frame fintech infrastructure in my head by looking at global spend (software, services, human capital) supporting traditional financial institutions (banks, insurance companies, etc.). It’s a rough heuristic as financial services will look different when encoded in software (did someone say crypto?), but the magnitude is mind-blowing. 

The lines between “SaaS company” and “API company” have blurred. In a previous era, there were three components of enterprise software & technology — business applications (thumbs up for system of record), middleware (yuck), and infrastructure (boxes and appliances). Many of the API-first poster children (Stripe, Twilio) initially were written off as middleware components. This pair grew to incredible heights over the last decade and paved the way for more to believe you could accrue value in that part of the stack. But, how would you characterize Stripe or Twilio today? Are they integration / infrastructure companies or applications that enable workflows? Somewhere in the middle probably. The lines demarcating an application and an API are blurring — workflow products are quickly releasing open endpoints after launch, and API companies are giving customers UIs and dashboards to deepen engagement with non-technical users. 

Dev tool and open source companies are priced to perfection, but I remain optimistic. Similar to the App Annie days following the first iOS explosion in the early 2010’s, investors are tracking projects across GitHub on all levels for any semblance of traction. It’s a funny rotation, because dev tools were considered a terrible place to invest up until a few years ago. The shift is a result of both organizational and technological forces. The increasing number of developers around the world have more sway within their companies to adopt new tools, and microservices & the cloud made it easier to apply a piece of technology to part of a system without overhauling lots of code. Since these projects are public in nature, the markets for them follow the information, and it’s hard to find value (in the traditional sense) as an investor. Yet, there are still dozens of unsolved problems that have now been put in developer’s hands making me bullish — observability & monitoring, tracing & debugging, security & networking, data transformation & analysis, deployment & orchestration, and much more. 

I spent more time this year learning about biology and life sciences. If I were to look out at the next several decades, it’s hard not to be excited about the potential of synthetic biology alone to reinvent therapeutics, food, materials, and more. Think about anything that requires some sort of chemical process and replace it with a biological one — the possibilities are endless. At first, it was easy for me to fall into the heuristics trap of “what’s the AWS of bio?” or “what’s the platform play here?” when organisms in the lab work nothing like bits and bytes. We can get closer to making it easily replicable through automation and machine learning, but the marginal cost of biological matter is fundamentally different from software’s. This realization has made me step back a bit from scrutinizing deeper sciences with the venture lens, but I continue to be energized about the future potential of what life sciences can bring us over the next few decades. 

The next few years may give birth to consumer Internet companies we can’t even begin to fathom. Predictions are hard. Yet, if applications and infrastructure continue their song and dance, it’ll be interesting to see how current build outs support new waves. These have mostly received attention for their ability to serve the enterprise, but how will they act as building blocks for entrepreneurs to build completely new services? From a big data ecosystem centered around Snowflake to an analytics universe with Databricks at the center to edge networks with Fastly and Cloudflare to the changing nature of the web with WASM or WebGL, the future looks promising.

Loading more posts…