The Long Hard Road To The Edge

A Year In The Life Of An Entrepeneur

1. July 2010: Ready: Set:  Delaware, the state with the lowest highest point. David Argentar, a biochemist by training and bioinformaticist by trade, has launched a startup. Of sorts. Well – more of a hobby, he’d be the first to admit. He has no business plan, no investors, no employees. All he really has, in fact, is an idea and a pending patent. And as everyone is eager to tell you these days, ideas are a dime a dozen, and patents are practically a scam.

It gets worse. Much. His idea is hardware. A new kind of solar concentrator, to be exact, made mostly of water. His first version was too heavy; but he thinks his redesign could conceivably, in his wildest dreams, drive down the cost of solar power by quite a lot. But—come on, now, really—a hardware startup? With only one founder?

Hardware is hard. It allows for no binary abstractions, no digitized purity to protect you from the real world. It is the real world, in all in its vicious and unforgiving glory, perpetually at the mercy of a hundred unexpected environmental factors. And almost by definition it is incredibly expensive to develop. I should know: I myself have a degree in electrical engineering – but I fled to the warm embrace of software as soon as I graduated. Hardware was much too temperamental for me.

Argentar, fortunately, is made of sterner stuff.

Good thing, too. Over the next year he’s going to need everything he’s got.

2. August 2010: Minions! Go! Argentar acquires two new things:
* a small crew of university undergraduates who have signed on for a work-experience term. They may not be experienced, but they only cost $5K/month.
* A blog. What follows is a selection of its greatest hits over the next roller-coaster year. (No, you can’t go read the whole thing: it’s mostly private. I have some access because I’m a friend of a friend.)

I have a tiny start-up, but I hesitate to call that my job, ’cause it doesn’t pay me anything, and in fact has an impressive burn rate. Since my patent applications will become public soon, I’m slowly lowering the veil of secrecy around my full-time hobby.

He and his minions get to work designing a working prototype of his concentrator. Everything seems to go well for a couple of months, until –

3. Early October 2010: Mo’ Hardware Mo’ Problems. They hit the first of many hardware pitfalls: parts. As Argentar says, “Badly written software often still runs, but badly build hardware usually doesn’t.”

Having now spent a long time on the phone with various prototype outfits, we’ve found that a couple of the ones we wanted to work with are offering very coarse tolerances. Both the rapid CNC-machining place and the one outfit that’ll do “rapid” extrusion (still a six-week lead time (because they have to make a die, then test it, then send us a sample part for approval, then adjust the die because the sample is wrong, then and so on and so forth)) are saying they can manage only ±0.01″/0.2mm on key dimensions. That’s too crude for optics work; ten-fold finer would be better.

and meanwhile

One of the other things we still need is a light source that can simulate full sun. I’m leaning towards LED arrays, and trying to get the overall power in the visible, NIR (meaning wavelengths < 1100nm, the bandgap of silicon solar cells), and far IR bands approximately correct. Problem: finding LEDs with longer wavelengths is difficult, and they’re not very efficient ( → very expensive) … The IEC is willing to let us use their new full-sun test cell… if we’re willing to wait until december, which, um, we’re not. *sigh* So, we still need to come up with some way of simulating sunlight.

I suspect that if I wasn’t captain of this ship, I’d be running for the lifeboats already. (The students can’t run; since they’re graded on this, they’re effectively chained to the benches.)

4. Late October 2010: The Scramble For Solutions.

The current plan for simulating sunlight is to use quartz-tube radiant heater elements combined with red and green LEDs to build up more or less the right spectrum. While the spectrum will be crudely approximated, this approach has the advantages that the heater bits are already long and thin, the LEDs in question are both cheap and efficient, and unlike using high-intensity IR LEDs to build up the IR spectrum, are unlikely to accidentally blind anybody. (Since I know academic labs are really sloppy about safety, I don’t want my team messing around with gear that can do them serious harm on my orders. I can’t help if they electrocute themselves with high-voltage power supplies, but I can avoid adding to the risk.)

Ah, another charming quirk of hardware: the threats of, say, accidental blinding or death by electrocution. Have I mentioned that I really like software?

The other problem is solved by the simple expedient of throwing his scarce money at it:

I’m going to buy a ~$15k mold, which probably won’t be useful after this prototyping stage. I can’t really afford this, but the other alternative would be to essentially abandon the student prototyping effort. I might be able to take the monetary loss on that, but the time is gone for good, and I’d still need to find a bunch of engineers to build a prototype for me. (That’s been previously estimated at $50-100k, and I certainly can’t afford that without investors, grants, or loans… but the point of having a prototype is to demonstrate that the technology is worth investing in.)

Writing off a $15k mold, but having labor at a fixed $5k/semester isn’t so bad… I could be flailing around while paying professional engineers $200+/hr each to help me learn in a particularly inefficient manner.

5. In Which Our Hero Considers The Lessons He Has Learned The Hard Way So Far

* I was even more unprepared for this than I suspected. Specifically, I should have done much more groundwork talking to vendors about tolerances and lead times.
* I’ve learned an expensive lesson about manufacturing tolerances and the difficulties of dealing with them. But I can’t see that there was anything that I could have done to prevent this, since I was too unfamiliar with the problem to grasp its difficulty.

6. November 2010: Plus/Minus

+ ordered optical components for prototype today
– critical path has no slack left
– discovered that bank silently reduced credit limit on company card, necessitating use of personal card for the Part of Unusual Complexity.
– still need to catch up on far too many things from ~9 week push on this project

7. Catastrophe Redux

Oh yes. Another endemic hardware problem: it breaks.

My student engineers and I have run into a little crippling problem with the prototype light concentrator array we’re building: the bare solar cells are too fragile to handle. We’ve broken most of ones we had trying to mount them, even handling them with tweezers. So, I’m looking for ideas on what to try next. I figure somebody who reads this must have some experience working with fragile things. …anybody got other ideas?

8. Into The Melee

And now, on to a milieu perhaps even less forgiving than hardware: business development.

I talked to a new lawyer (at the same law firm) about the contract I want to sign with the current /bus/dev/ folks. It turns out the reason my patent attorney wanted me to talk to him is that he tried to start a company, but was betrayed by one of the hired suits they got to run it, which caused a funding deal to collapse.

So… he had lots of skeptical things to say about /bus/dev/ companies and in fact anybody else who tries to get money out of start-ups. In short: caveat emptor, and keep your laser legal team handy. and now I’ve got a long list of questions to ask the /bus/dev/ guys about the contract and a another measure of circumspection to add to my deep suspicion of the /bus/dev/industry.

iIm also clear that the fundamental relationship I’m entering w/ the /bus/dev/ guys is prospector to over-priced-equipment vendor during a gold rush. While lots of people talk about “long term relationships”, it’s fundamentally true that the /bus/dev/ guys are paid in cash, and if my company flames out, it will at worst just disappoint them.

9. December 2010: Victory Snatched From The Jaws Of Defeat?

It appears the solution to the strengthening and handling question is double-sided tape. It’s cheap, it’s off-the-shelf with no lead time, and it’s relatively strong… The cells we have wired up are working very close to perfectly, which is a big relief.

As expected, our construction schedule has gone *poof*! one vendor — who shall remain nameless so long as they don’t further annoy me — finally shipped parts, only two and a half weeks late. But we now have the parts and test equipment in hand. That sounds lame, but it’s a surprisingly big step: it means everything could be manufactured with very short timelines, and was.

Most importantly, the quality of the optical components is far beyond what I was expecting … this means that an extremely cheap process can produce a difficult part for pennies a square meter. My fantasies involving lopping two-thirds off the price of solar power seem much less fantastic now.

I’ve just signed the contract hiring a business development firm. Because of the way it’s set up, I’m up on the hook for at least $16k over the next two months. Even though I’m fully aware of the cost of doing business as a high-tech start-up, that’s still bracing.

10. Victory!

Q: what is the fully amortized cost of a watt of electricity generated from coal?

A: $2.22

Q: what is the fully amortized cost of a watt of electricity generated by a short production run of my panels?

A: $2.17, according to my student engineers. I have not verified this number, nor do I know what assumptions went into it. I know they haven’t thought too hard about it, because they got the production cost estimates just this morning, so I suspect there may be some error.

However, this is for the existing design, which is just a step up from bear skins and stone knives… and a longer production run gets the cost down to around $1.50/W, according to the students. This is strongly suggestive that with a better design, some cost engineering, and a production run of hundreds of thousands of panels, I’ll be able to reach the mythical $1/W cost, and dominate the world power industry (insert maniacal laughter here).

11. …maybe.

Revised A: $2.58

That’s a very rough cost of manufacturing one of my panels to produce a watt of power at peak rating. It’s not as useful a number as I’d hoped, but still isn’t bad either. It’s deliberately a high estimate, since where we had to fudge extrapolate numbers, we aimed high, so that all the error is against us. For example, we used the retail cost we paid for a single box of 12 solar cells to estimate the cost of silicon in the system. Just accounting for regular wholesale markup should roughly halve that, and bulk orders from solar cell foundries might reduce it to a tenth that.

Since plastic and metal costs were roughly equal to the cost of silicon in our spreadsheet, that knocks the guesstimated peak watt cost to a buck and change, easily enough to undercut the grid. Go us! I’ll eventually have to get a total cost of ownership, but looking at other vendors’ claims suggests that will put me ahead of the game — nobody wants to own up to that.

12. January 2011: The Patent Game

An international patent examiner has looked over the prior art and found a patent he claims makes my work obvious. Which would be a crippling problem except for the similarities ending with the words “light concentrator for solar energy use”. As it is, it’s just a nuisance: I know about this patent, my attorney knows about this patent, and we know how to overcome such claims.

I don’t want to say the differences should be obvious on inspection, but they’re very easy to state, largely involving mechanical complexity and cost of the prior art. Oh, and just some utterly trivial differences in the way it functions. I mean, a diesel engine is just like a steam engine, right? Y’know: pistons, heat, intake, exhaust, motion, blah, blah, blah.

Legally, inventors are expected to know the entire prior art in the field they’re inventing in. Inventors are also expected to cite all relevant prior art in the patent application, although the law punishes only deliberate omission. The main practical effect of having cited it is that we also listed a few reasons it doesn’t apply, so later, binding examinations can see that we were already aware of the prior art, and give our objections more credence.

Apparently, I got off easy. It’s unusual that this round of patent examining finds only a single pre-empting patent. This is a very good sign that my patent application will succeed, and gives me hope that it will issue quickly.

13. The Business Game

The business plan has been declared done. What we ended up with sure looks like a best-seller: even allowing for huge mark-ups and large, unanticipated costs, we think that my solar arrays can undercut the power grid with another year’s worth of engineering, and a year after that the company will turn a profit of a half million dollars. Growth thereafter is limited only by our ability to raise capital to keep up with demand.

Now, everybody who deals with business knows this is genre fiction, but one of the little white lies of the business world is that we treat it as fact.

14. February 2011: Running To Stand Still

My to-do list grows without bound, which is a fancy way of saying I’m overwhelmed. There are simply more things to do than I have time for (at least while preserving what’s left of my sanity). Bringing investment in will help some, but too many of the things seem to require me. While gathering data tomorrow, I’m not — for example — working on my pitch for the upcoming poster session. While the /bus/dev/ guys can do a lot of that stuff for me, I have to polish the result for my speaking style. And I need to chase after a particular business contact who the /bus/dev/ guys don’t know yet. And so on.

15. March 2011: The Rubber Chicken Circuit


I talked to a local angel/VC fund. Only three guys from the fund there: the head, his lieutenant, and their minion. They seemed very interested, and asked lots of questions. The head even called up another VC that only does energy plays, and gave them my pitch(!). Other VC interested: asked detailed technical questions, wants to see business plan.

Result: their lips said no, but the way the /bus/dev/ guy interprets their answer is that they want the energy VC (or possibly somebody else) to take the lead, and they’ll follow, because they lack the technical background to really evaluate my technology. The /bus/dev/ guy thinks that if the energy VC is interested, the two funds will will come up with all the money I need to get stuff on the market, ~$850k, probably split $600k/$250k. /bus/dev/ thinks this is the sort of deal they want because they asked for our nominal valuation ($3.25 million, by the pneumatic valuation method, ie we pulled a number out of the air), and didn’t balk when we said that the $850k would get them about a quarter of the company. (We don’t think they’ll take that offer, but merely accept that it’s our opening position.)


This group was a large investor consortium; there were 33 people at the table, and some unknown number watching via video conference at remote sites. They were interested, but asked a bunch of questions I didn’t have answers to, mostly about money. Fast, definitive turn-around, though: they voted while-I-waited. Answer: no. The claimed reason is that they look at later stage companies. We suspect they’re concerned about technical risk.


An investment fund said no. I knew this pitch wasn’t gonna go well as soon as they said their investors are union pension funds. While they’re looking for risk to drive up returns, they’re not looking for that much risk. The feedback we got was worth the nickel, though. They like the technology, but want to see some operating experience. That’s the same as we’d gotten from the previous bunch who were unimpressed.

16. May 2011: Seeded!

I now have

1) a logo
2) investors
3) a salary

which means this quixotic quest goes from being an expensive hobby to being a job. My salary is charmingly low, but its real point is just to make it possible for me to continue to play tech entrepreneur until Serious Money shows up, which ought to happen after the next prototype has sat on peoples’ roofs for a few months.

17. July 2011: Preparing For Liftoff

Need to round up engineers soon. Way behind on that as a consequence of being way behind on working up my data. To paraphrase an ancient videogame, “startup geek needs staff, badly.”

Today, Sun Edge has $200,000 in seed funding, and a new batch of students is working on the concentrator’s frame and cooling system, while professionals are being brought in to build the next, fully functional prototype, with a target date of early 2012. Argentar has already sunk a year of his life and $100,000 of his own money into the project … but even though the science still seems very promising, there’s no guarantee of market success. An entrepeneur’s life is never easy. When he or she is building hardware, it’s even harder.