The five technical challenges Cerebras overcame in building the first trillion transistor chip

Superlatives abound at Cerebras, the until-today stealthy next-generation silicon chip company looking to make training a deep learning model as quick as buying toothpaste from Amazon. Launching after almost three years of quiet development, Cerebras introduced its new chip today — and it is a doozy. The “Wafer Scale Engine” is 1.2 trillion transistors (the most ever), 46,225 square millimeters (the largest ever), and includes 18 gigabytes of on-chip memory (the most of any chip on the market today) and 400,000 processing cores (guess the superlative).

CS Wafer Keyboard Comparison

Cerebras’ Wafer Scale Engine is larger than a typical Mac keyboard (via Cerebras Systems)

It’s made a big splash here at Stanford University at the Hot Chips conference, one of the silicon industry’s big confabs for product introductions and roadmaps, with various levels of oohs and aahs among attendees. You can read more about the chip from Tiernan Ray at Fortune and read the white paper from Cerebras itself.

Superlatives aside though, the technical challenges that Cerebras had to overcome to reach this milestone I think is the more interesting story here. I sat down with founder and CEO Andrew Feldman this afternoon to discuss what his 173 engineers have been building quietly just down the street here these past few years with $112 million in venture capital funding from Benchmark and others.

Going big means nothing but challenges

First, a quick background on how the chips that power your phones and computers get made. Fabs like TSMC take standard-sized silicon wafers and divide them into individual chips by using light to etch the transistors into the chip. Wafers are circles and chips are squares, and so there is some basic geometry involved in subdividing that circle into a clear array of individual chips.

One big challenge in this lithography process is that errors can creep into the manufacturing process, requiring extensive testing to verify quality and forcing fabs to throw away poorly performing chips. The smaller and more compact the chip, the less likely any individual chip will be inoperative, and the higher the yield for the fab. Higher yield equals higher profits.

Cerebras throws out the idea of etching a bunch of individual chips onto a single wafer in lieu of just using the whole wafer itself as one gigantic chip. That allows all of those individual cores to connect with one another directly — vastly speeding up the critical feedback loops used in deep learning algorithms — but comes at the cost of huge manufacturing and design challenges to create and manage these chips.

CS Wafer Sean

Cerebras’ technical architecture and design was led by co-founder Sean Lie. Feldman and Lie worked together on a previous startup called SeaMicro, which sold to AMD in 2012 for $334 million. (Via Cerebras Systems)

The first challenge the team ran into according to Feldman was handling communication across the “scribe lines.” While Cerebras chip encompasses a full wafer, today’s lithography equipment still has to act like there are individual chips being etched into the silicon wafer. So the company had to invent new techniques to allow each of those individual chips to communicate with each other across the whole wafer. Working with TSMC, they not only invented new channels for communication, but also had to write new software to handle chips with trillion plus transistors.

The second challenge was yield. With a chip covering an entire silicon wafer, a single imperfection in the etching of that wafer could render the entire chip inoperative. This has been the block for decades on whole wafer technology: due to the laws of physics, it is essentially impossible to etch a trillion transistors with perfect accuracy repeatedly.

Cerebras approached the problem using redundancy by adding extra cores throughout the chip that would be used as backup in the event that an error appeared in that core’s neighborhood on the wafer. “You have to hold only 1%, 1.5% of these guys aside,” Feldman explained to me. Leaving extra cores allows the chip to essentially self-heal, routing around the lithography error and making a whole wafer silicon chip viable.

Entering uncharted territory in chip design

Those first two challenges — communicating across the scribe lines between chips and handling yield — have flummoxed chip designers studying whole wafer chips for decades. But they were known problems, and Feldman said that they were actually easier to solve that expected by re-approaching them using modern tools.

He likens the challenge though to climbing Mount Everest. “It’s like the first set of guys failed to climb Mount Everest, they said, ‘Shit, that first part is really hard.’ And then the next set came along and said ‘That shit was nothing. That last hundred yards, that’s a problem.’”

And indeed, the toughest challenges according to Feldman for Cerebras were the next three, since no other chip designer had gotten past the scribe line communication and yield challenges to actually find what happened next.

The third challenge Cerebras confronted was handling thermal expansion. Chips get extremely hot in operation, but different materials expand at different rates. That means the connectors tethering a chip to its motherboard also need to thermally expand at precisely the same rate lest cracks develop between the two.

Feldman said that “How do you get a connector that can withstand [that]? Nobody had ever done that before, [and so] we had to invent a material. So we have PhDs in material science, [and] we had to invent a material that could absorb some of that difference.”

Once a chip is manufactured, it needs to be tested and packaged for shipment to original equipment manufacturers (OEMs) who add the chips into the products used by end customers (whether data centers or consumer laptops). There is a challenge though: absolutely nothing on the market is designed to handle a whole-wafer chip.

CS Wafer Inspection

Cerebras designed its own testing and packaging system to handle its chip (Via Cerebras Systems)

“How on earth do you package it? Well, the answer is you invent a lot of shit. That is the truth. Nobody had a printed circuit board this size. Nobody had connectors. Nobody had a cold plate. Nobody had tools. Nobody had tools to align them. Nobody had tools to handle them. Nobody had any software to test,” Feldman explained. “And so we have designed this whole manufacturing flow, because nobody has ever done it.” Cerebras’ technology is much more than just the chip it sells — it also includes all of the associated machinery required to actually manufacture and package those chips.

Finally, all that processing power in one chip requires immense power and cooling. Cerebras’ chip uses 15 kilowatts of power to operate — a prodigious amount of power for an individual chip, although relatively comparable to a modern-sized AI cluster. All that power also needs to be cooled, and Cerebras had to design a new way to deliver both for such a large chip.

It essentially approached the problem by turning the chip on its side, in what Feldman called “using the Z-dimension.” The idea was that rather than trying to move power and cooling horizontally across the chip as is traditional, power and cooling are delivered vertically at all points across the chip, ensuring even and consistent access to both.

And so, those were the next three challenges — thermal expansion, packaging, and power/cooling — that the company has worked around-the-clock to deliver these past few years.

From theory to reality

Cerebras has a demo chip (I saw one, and yes, it is roughly the size of my head), and it has started to deliver prototypes to customers according to reports. The big challenge though as with all new chips is scaling production to meet customer demand.

For Cerebras, the situation is a bit unusual. Since it places so much computing power on one wafer, customers don’t necessarily need to buy dozens or hundreds of chips and stitch them together to create a compute cluster. Instead, they may only need a handful of Cerebras chips for their deep-learning needs. The company’s next major phase is to reach scale and ensure a steady delivery of its chips, which it packages as a whole system “appliance” that also includes its proprietary cooling technology.

Expect to hear more details of Cerebras technology in the coming months, particularly as the fight over the future of deep learning processing workflows continues to heat up.

Voyage’s driverless future, ghost work, B2B growth strategies, and Black Hat takeaways

Inside Voyage’s plan to deliver a driverless future

In the autonomous vehicle space, startups have taken radically different strategies to building our AV future. Some companies like Waymo have driven all across different types of environments in order to rack up the datasets that they believe will be needed to effectively maneuver without a human driver.

That’s the opposite strategy of Voyage, where CEO and founder Oliver Cameron and his team have focused on driving safety in the incredibly constrained context of two retirement communities.

Our transportation editor Kirsten Korosec talked with the company and analyzes their approach in a new profile for Extra Crunch, and also drops some news about a partnership the company has brewing with a major automotive manufacturer.

Cameron, who shies away from discussing timelines, describes the company as inching toward driverless service.

Its self-driving software has now reached maturation in the communities it is testing in, and Voyage is now focusing on validation, according to Cameron.

Voyage has developed a few systems that will help push it closer to a commercial driverless service while maintaining safety, such as a collision mitigation system that it calls Rango, an internal nickname inspired by the 2011 computer-animated Western action-comedy about a chameleon.

This collision mitigation system is designed to be extremely fast-reacting, like a reptile — hence the Rango name. Rango, which has an independent power source and compute system and uses a different approach to perception than the main self-driving system, is designed to react quickly. If needed, it will engage the full force of the brakes.

Startup ads are taking over the subway

Public transit is just swimming in startup ads. From complete Brex takeovers of the San Francisco Caltrain station to the sleep puzzles posted by Casper across the New York City subway, startups have been taking advantage of this unique out-of-home advertising space. What’s the full story though? Our reporter Anthony Ha takes a look at how the subway ad market came to be in the past few years, and what the future holds for other marketers.

WeWork S-1, building marketplaces, improving content marketing, and the demise of Tumblr

WeWork’s S-1 misses these three key points

After much discussion, WeWork finally dropped its S-1 filing with the SEC today as it makes preparations for its IPO. While the company has been producing sizable revenues the past few years, the company didn’t disclose everything I think it needed to in order for investors to make a judgment about its financial future.

It’s not as though WeWork hasn’t tried to give us some insight in its S-1. One of WeWork’s core operating metrics is “contribution margin including non-cash GAAP straight-line lease cost” (or what I will abbreviate just this one time as CMINCGAAAPSLLC). Through this metric, the company offers us a single number into the health of its business — essentially a way for investors to understand the performance of the company’s mature office locations.


What’s missing here though is that WeWork has aggregated its finances for hundreds of locations down to a summary statistic, complemented with a huge amount of text devoted to describing the evolution of a property from lease signing to mature profit-making office. At no time does the company describe the contribution margin and how it changes throughout the course of a single lease. Instead, it provides the following completely numbers-free chart showing that … it makes more money as time goes on.

How even the best marketplace startups get paralyzed

Marketplaces are hard to build. You have to generate both supply and demand, and if that isn’t bad enough, you then have to work to match both sides of the marketplace to get a transaction to clear (and therefore generate revenue).

WeWork’s S-1 misses these three key points

No startup is as polarizing as WeWork, and for good reason. The company, whose relentless growth has seen it open 528 locations across 111 cities in just about nine years, has never been entirely forthcoming on exactly how the unit economics add up at its locations. And so we have had a beautiful Rorschach test for the financial class these past few years regarding the company: it’s either the greatest financial return of all time or a Ponzi scheme (and absolutely nothing  in between dammit).

That ambiguity is supposed to change with the company’s S-1, where it is required by law to show a reasonably comprehensive set of numbers to investors in order to go public. Unfortunately, despite all the verbiage (“Our mission is to elevate the world’s consciousness.”) and data, we still don’t know the health of the core of the company’s business model or fully understand the risks it is undertaking. 

Here are three questions that remain unanswered so far by the company’s filing.

No cohort data on contribution margin

As I pointed out a couple of months ago, the ability for investors to understand the true unit economics of WeWork’s business is critical for cutting through the debate over its financial future.

It’s not as though WeWork hasn’t tried to give us some insight in its S-1. One of WeWork’s core operating metrics is “contribution margin including non-cash GAAP straight-line lease cost” (or what I will abbreviate just this one time as CMINCGAAAPSLLC). Through this metric, the company offers us a single number into the health of its business — essentially a way for investors to understand the performance of the company’s mature office locations.

How young VCs bootstrap new venture firms

We spend a lot of time talking about new funds, and new startup venture raises, but we spend little time talking about the cash flow challenges of running a venture fund. Let’s change that today.

Starting a new venture fund is extremely challenging. In addition to just the monstrous task of fundraising — which can take as long as two years in some cases to lock down all the limited partners (LPs) on the same terms — the economics for a debut fund are often just terrible.

Take a sort of starter $20 million seed fund with two general partners using the industry’s oft-quoted (but not really all that common) “2 & 20” compensation model. This hypothetical fund rakes in $400,000 a year in management fees (2% of $20 million) to cover all costs of the fund: office rent, staff costs, legal fees, tax preparation, and accounting services in addition to the travel and entertainment costs of trying to woo founders. Whatever remains is split between those two GPs as their salaries. It’s not uncommon for new partners to make $50k — or even nothing — in the early years of a new firm, which is one reason the industry is stacked with ultra-wealthy individuals.

For young financiers looking to break into the industry, the situation is bleak, which is one reason why fund managers have gotten very creative around how to structure their management fees in order to bootstrap a venture firm in its early years.

These sorts of fund details are often kept tight-lipped, but thanks to the Mike Rothenberg case, we now actually have real data from a new firm and how it structured its fees for asset growth. From discussions with others in the industry, the models that Rothenberg Ventures used are reasonably available for investment managers looking to build new franchises.

All data for this analysis comes from Exhibit A — the Expert Report of Gerald T. Fujimoto, a forensic accountant who evaluated Rothenberg Ventures as part of the SEC’s lawsuit against Mike Rothenberg (Case No. 3:18-cv-05080). The exhibit was filed July 29, 2019. TechCrunch did not attempt to verify the work of the forensic accountant, since this analysis makes no claims about Rothenberg Ventures, but uses the data for illustrating how funds are structured in today’s work.

Below is a recreation of the fund structures as reported in the SEC’s case against Mike Rothenberg. Rothenberg Ventures raised a series of four venture funds with fee structures that vary widely from the traditional 2 & 20 model, which assumes a 2% annual management fee for each of the ten years of a fund’s life (further extensions beyond ten years don’t usually offer significant fees, although every fund is structured differently). That equation means that management fees generally represent 20% of a fund’s committed capital.

sec v michael rothenberg exhibit a

Source: SEC v. Michael Rothenberg (Exhibit A)

For the firm’s debut fund, Rothenberg entirely eliminated the slow and orderly parceling out of fees in lieu of a one-time 17.75% fee upon the closing of the $2.6 million fund. That meant an immediate infusion of about $470,000 into the firm, but no continual fees thereafter. This sort of heavy upfront payment is not uncommon in the industry, although it is less common to have literally the sum total of management fees for a 10-year fund paid out entirely on its first day.

From an LP perspective, this sort of fee structure indicates that the firm almost certainly would have had to raise additional funds almost as soon as the first one closed, since the fees of future venture funds would be needed to cover the management costs of the first fund in its later years.

In short: this is what a bootstrap looks like in venture capital.

Now, continuing to the second fund (2014), we see a bit more of a traditional parceling out of fees over the course of the fund, although still with a heavy upfront skew. The fund pays out the typical 20% of invested capital in total, but 80% of that amount was paid out in the first two years. Again, the implicit assumption with this sort of bootstrap is that the firm will succeed and raise additional capital (and therefore management fees) to keep the operation going.

We then see the same pattern in the 2015 fund, with fees having a normal structure, but then with more aggressive upfront payouts required. So while the fund had a flat payout every year for its management fee, two years of that fee was to be paid out immediately upon close. Similarly, the administrative fee was flat — but paid out entirely in the first year of the firm’s operation.

Finally, the fourth fund (2016) returns to a more typical, flat fee structure at 2.5% per year with no provisions for upfront payment.

Why does this all matter? Let’s go through a back-of-the-napkin exercise of what these numbers really meant for the operations of the fund:

copy kobalt streaming service total maus paid maus

Data from SEC Case, Exhibit A

As we can clearly see here, all of those management fees upfront really did give the firm far more resources in the early years than it might have otherwise had. Over the first three years, the firm had access to roughly $5.1 million in fees, whereas with a traditional 2% annual structure, the firm would have had access to just $1.2 million. Of course, that bootstrap comes at a cost in the later years, when the firm would have more resources to manage the fund.

Nonetheless, those upfront payments helped the firm tremendously punch above its own weight. With its $1.2 million of fees in year one, it essentially had the resources of a $60 million fund — yet it had only raised $6.7 million. It similarly punched above its weight the next few years as it raised new funds with aggressive upfront fee schedules as well.

Of course, there’s a heavy burden with this approach — it’s a bet-it-all strategy that leaves little room for error (such as a series of failed investments) that might make future fundraising hard. It’s a rocket with no ejection seat, but when it works, it can compress the time to venture fund leader dramatically — maybe even by as much as a decade.

Ultimately, VCs like to bet, and they certainly like to bet on themselves, which is why these sorts of cash-flow optimizations are prevalent for new firms. No one assumes that their firm is going to fail. Plus, these sorts of management fee structures are also among the few tools a non-wealthy individual can use to even get a new fund underway. For new fund managers and for others considering jumping into the VC industry, a nuanced understanding of the risks and opportunities of mortgaging future dollars for present spend is critical — not only for one’s integrity and stress levels, but hopefully to avoid those SEC investigators and forensic accountants as well.

Kobalt, Apple and smartwatches, Hadoop, customer support, and social work and AI

The Kobalt EC-1: How a Swedish saxophonist built Kobalt, the world’s next music unicorn

My favorite pieces we host on Extra Crunch are our EC-1 series of in-depth profiles and analyses of high-flying, fascinating startups. We launched Extra Crunch with a multi-part series on Patreon, and then we covered augmented reality and Pokémon Go creator Niantic and gaming platform Roblox.

This week, Extra Crunch media columnist Eric Peckham launched the first part of his three-part EC-1 series looking at music “operating system” startup Kobalt. Kobalt is not perhaps a popular household name like Roblox, but it’s influence is heard pretty much every single time you listen to music. Kobalt is upending the traditional infrastructure to track music plays to capture royalties for artists, an industry that today still involves people literally walking into bars and writing down what’s playing. From that base, Kobalt wants to expand into services to empower the next-generation of stars and mid-market talent.

What I loved about this story is that not only is Kobalt completely rebuilding an otherwise stagnant industry, but its founder and CEO is also such a dynamic individual. Willard Ahdritz was a former saxophonist whose band was essentially abandoned by their music label — even while that label wouldn’t give up the economics that would allow the band to continue (some founders may have similar experiences with their venture investors). Ahdritz would eventually start his own music label called Telegram, and a bit later started Kobalt to solve the problems he kept running into on the music publishing side.

It’s been almost two decades, but today, Kobalt offers a suite of technologies and services and has its crosshairs on the big three labels — Universal, Sony, and Warner. It’s also raised a boatload of venture capital and is closing in on a unicorn valuation. Read the full story, learn more about this analytically fascinating business, and get ready for parts two and three coming soon.

Refer a friend to Extra Crunch

HealthTech VCs, fundraising in August, reducing churn, North, and co-ops as startups

What tech gets right about healthcare

This week, our long-time healthtech correspondent Sarah Buhr href="">talked to leading health VCs Phin Barnes of First Round Capital, Matt Ocko of DCVC, and Nick Naclerio of Illumina Ventures about what they are seeing in the healthtech ecosystem, how they are thinking about investments in the space, as well as the reasons behind why they led their recent deals in health startups.

[DCVC’s] thesis is simple: if the cost for superior, life-saving care is half to 10x less, and results are 10-100 times better, then adoption happens quickly, and hospitals and insurance companies are hard-pressed to say no to saving money while lives are saved. Some of the healthcare companies we have invested in have achieved dramatic results to help deliver “disruption from within”.

One example is Karius, which uses genomics and AI to advance infectious disease diagnostics. The company can recognize almost every pathogen mankind has ever encountered at the genomic level.

Using machine learning algorithms that allow for rapid analysis of complex genomic data, they are pioneering new testing methodologies to accurately detect and characterize infectious diseases. Instead of relying on a century-old method developed by Louis Pasteur that can take weeks for a result, the Karius gene-based blood test that can return results in a day.

This can make the difference between life and death. And Karius’ technology is backed by multiple large-scale studies, peer-reviewed publications, and a delivery and reimbursement model that satisfies both doctors and administrators in the existing framework of the healthcare system.

How to fundraise in August

Fundraising is always brutal (even if journalists never cover the gritty and gory details). And August would seemingly be the most brutal month to fundraise, what with everyone on vacation. The reality though is that while you can’t fundraise in August the way you might in September or October, there are a lot of strategies you can use to take advantage of the uniquely relaxed tenor that August offers. I provide some context for how to fundraise in August, as well as some tactics to implement to maximize your fundraising efficiency in the fall.

How to fundraise in August

August is often considered the black hole of venture capital fundraising. Everyone is on vacation (well, everyone who’s not a founder anyway), while half of Silicon Valley is slogging down to Black Rock City for Burning Man. It understandably can just seem like an exercise in futility to try to raise any funding at all.

I’m here to tell you though that August is not the bleakest month of the year for fundraising (that actually would be December according to data from DocSend we’ve published). In fact, using August effectively for fundraising is perhaps the single most important factor for success in the coming fundraising season (there is a reason that YC Demo Day, one of the largest fundraising events in the calendar, is set for August 19-20 after all).

Let’s walk through a plan of attack.

First, the truth about VCs and vacation

Let’s get one thing out of the way: Yes, VCs take vacation, sometimes sparklingly expensive ones, like the kinds with yachts or the kinds where someone rents out a whole ski chalet (or two). It can seem like an incredibly enviable lifestyle, and it is at a certain point of success, particularly in comparison to the context of a founder who is working around the clock and eating instant ramen.

Morty raises $8.5M series A to help first-time homebuyers secure their mortgages

For the past decade, Brian Faux has been fighting on the front lines of housing finance. In between pursuing a career in mortgage lending and holding stints at Freddie Mac and Wells Fargo, Faux spent more than two years in the detritus of the 2008 financial crisis advising the Department of Housing and Urban Development on how to recover the housing markets through the creation of the Distressed Asset Stabilization Program.

Now Faux, along with co-founders Nora Apsel and Adam Rothblatt, is working to take those hard-learned lessons and build a streamlined and simple mortgage broker online, particularly for first-time homebuyers. Through New York City-based Morty, the trio and their team have launched a tool that allows homebuyers to understand exactly what their buying power is and which homes they can afford.

That product has captured the attention of investors. The company announced today an $8.5 million Series A fundraise led by Prudence Holdings, with participation from Lerer Hippeau and Thrive Capital, the firm which had led Morty’s seed round in 2017. Prudence, a family office managed by Gavin Myers, previously backed real estate brokering startup Compass, and the Morty team first met the firm through participation in TechStars New York.

Morty’s main product guides homebuyers through the process of getting mortgage pre-approval and then finding and signing a loan with a mortgage lender. Through a “Home Financing Score,” the platform visually breaks down the factors that can lead to approval or rejection of a mortgage application, allowing users to optimize their finances to maximize their buying power.

While code operates much of the underwriting and origination process, there is a human touch as well. Faux explained that with current mortgage options, “It’s still too scary. It’s still too opaque, [so consumers] want that human interaction, eventually, but they just wanted it on their terms. And nobody’s kind of brought that to them” before Morty.

Apsel said that “As well as having a digital platform that automatically verifies and underwrites people so that they know exactly how much they qualify for, we also have have mortgage experts on staff available to help people through every step of the home buying process.”

She says that transparency and education have been key to Morty’s early indicators of success. “What we have found is that as long as you are communicating those things to all of the necessary parties — the homebuyer, the realtor, the title agent, everybody — it works. It’s the lack of transparency, and it’s the lack of communication that I think has frustrated this industry for so long,“ she said.

Z9A9266 FINAL 2 updated cleaner white background 1

Morty founders Adam Rothblatt, Brian Faux, and Nora Apsel. Photo via Morty

Morty, which at launch had licenses to operate in 10 states, has now expanded to cover 34 states. One notable exception though is New York, which has particularly stringent and slow-moving licensing processes. The company is hoping to have full nationwide coverage in the years ahead.

Faux says that while the startup focuses on first-time homebuyers, there is nothing preventing the company from expanding to repeat home sales as well. “Once you build trust, and once you show them who you truly are, unbiased and just looking out for their well being, they’ll come back to you,” he said.

Startups related to home buying have received intense attention from investors, with companies like Blend and Opendoor receiving nine-figure infusions of capital over the past few weeks. And Morty is also up against incumbents like LendingTree, which aggregates loans in a variety of categories. Morty’s differentiation is ultimately its focus on ease-of-use, as well as its wide licensing.

Series A(ggregate)

We spend a lot of time around here covering the latest startup fundraises, and for good reason. While capital is certainly an input and not an output, there is nothing quite like the closing of a round of several million in venture capital to prove that yes, the startup I’m working on is at least interesting to someone other than me. External validation shouldn’t be your motivating principle, but it is motivating. Plus, it’s a great milestone to reach out to the press and start talking up the story.

And so week after week, we cover the latest rounds. This company raised $4.5 million in a seed round, and this company raised $16 million in a series A. These stories — and the narratives behind them — are crisp, clean, and precise. A proverbial founder walked up and down South Park in SoMa, explained their story, collected a couple of term sheets, picked one, locked in the due diligence, and is now announcing their round. The VCs are excited, the founder(s) are excited, the employees are excited (and sometimes even the customers are excited!)

The reality for founders though is far more messy and gritty than those headlines would indicate. When I get founders off the record and out for drinks, the true story starts to emerge. That $4.5 million seed fundraise took eight months of maniacal scheduling with two hundred investors just to find a lead. And that lead didn’t lead lead, but took only 20% of the round. In the meantime, they raised twelve times across convertible notes and SAFEs, each one giving the company just a bit more gas in the tank to continue.

When I wrote that a startup raised $4.5 million in one slam dunk, what I really should have written was that they raised $150k, $300k, a few more $50k investments from randos, a couple of thousand from that startup competition, wow $500k from that amazing angel, a $750k SBIR grant from the government that took nine months too long to process, some credits from Brex, and finally at some point that lead investor showed up who gets $3-3.5 million in news value credit on their wimpy $900k check.

As an editor and a writer who covers these aggregate rounds, I struggle with how to approach them. Founders regularly tell me that they would love more transparency and less bravado around fundraises. They want to read how other founders handle the messy complexity of their fundraises, if only because they can compare their own hellish experiences with those of others.

More fundamentally, our readers deserve to read the truth. A $4.5 million round led by a single venture firm writing a $3.5 million check is a very different construct than a bricolage of a random assortment of angel investors. That difference in investor and round quality does indicate something about the startup under examination, and so offering more of those details would better inform our readers as well.

All that is well and good, but no one really wants to hear about these difficulties. Certainly users and customers don’t want to hear about how the software they use or purchased is run by a company that is constantly days away from death. Some early-stage employees probably have the focus to ignore such morbid considerations while carrying out their functions, but many need their paychecks to come from a black box. Somehow, the checks always arrive, and that lowers the stress for everyone.

And even just in terms of the craft of writing, do we really want to exchange the standard funding sentence (“blah blah blah raised blah from blah with participation from blah blah blah”) with a multi-paragraph exegesis of a fundraise?

Writing is about choosing which details are salient and which to pass over. It would be exhausting every morning to read tomes of fundraise detail. Yet, our consistency in depicting fundraises as efficient and precise can create an atmosphere where if you didn’t find a lead in a few weeks and lock down the whole round, you are a failure.

That’s not really a depiction I want to support.

And so, take this as someone who talks to dozens of founders a year off the record about their fundraises, and also sat on the other side of the table as a VC for years. Fundraises are almost always really, really, tough. Very few people get commits in the first meeting, or even in the subsequent meetings. Half the investor introductions during a fundraise are often a complete waste of time if not outright damaging, psychologically or materially. There are a lot of sharks out there. It is much more common today to aggregate a bunch of mini-rounds than it was a couple of years ago.

This is not failure, but just the path of the entrepreneur today in 2019. And at the end of that whole long and windy road, after all of those hundreds of hours of coffee meetings and PowerPoint strategy sessions and skeptical investor convos, all of that work will boil down to twenty words about how the fundraise closed, X dollars were raised, and money was seemingly wired magically to your bank account.

You, me, and really everyone can and should know the truth. But perhaps just rejoice in that headline, and get back to the next slog.