How Ama­zon changed start-up fund­ing re­quire­ments


Start-ups gob­ble cap­i­tal. In the tech space a lot of the fund­ing re­quire­ments used to be for server in­fra­struc­ture, which had to be bought. Given how hard it is to pre­dict de­mand or the ac­tual time it takes to build an on­line busi­ness, this meant that most businesses ei­ther over-or un­der­in­vested in tech, and a lot of cap­i­tal was wasted through over-ca­pac­ity and de­pre­ci­a­tion.

When Ama­ launched Ama­zon Web Ser­vices in 2006, it changed the amount of f und­ing re­quired for tech start-ups be­cause it could now lease in­fra­struc­ture by the hour. The re­sult is that start-ups are far cheaper than they used to be and far more ideas can be tested in the mar­ket for a lot less cap­i­tal at risk. Fund­ing re­quire­ments have come down as a re­sult, and a smaller fund can now have as much im­pact as a much larger fund used to.


In 1999/2000 I worked as a ven­ture cap­i­tal­ist in Lon­don, in­vest­ing in early -stage In­ter­net deals. Aside from the nor­mal chal­lenges of iden­ti­fy­ing which ideas could be turned by the teams that came with them into bil­lion- dol­lar businesses, the tech space at the time was also char­ac­terised by real prob­lems in terms of the as­sets re­quired to ac­tu­ally run an on­line busi­ness, and the cap­i­tal re­quired to fund these.

You see, back in the dark ages, if you wanted to run a web­site you needed to own the phys­i­cal in­fra­struc­ture. This meant that you needed your own web­servers, load-bal­ancers, air-con­di­tioned

server rooms, RAID stor­age and con­nec­tion to the main In­ter­net l ines. Ev­ery­thing was or­dered, paid for, de­liv­ered, in­stalled, tested and us­able.

The cru­cial and com­plex na­ture of the in­fra­struc­ture meant that the star­tups of­ten had their best de­vel­op­ers set­ting up and test­ing servers. A sin­gle server could take six weeks to ar­rive and t wo days just to get up and run­ning. It also meant that the work­ing cap­i­tal needs of the busi­ness were far greater than they would be for the equiv­a­lent busi­ness to­day, and so much harder to pre­dict.


As any­one who has ever tried to build an app, e-com­merce site, or run a big tech project will tell you, it’s very hard to pre­dict how long it will take to build. In­ter­de­pen­den­cies in con­nected sys­tems al­ways throw up un­ex­pected glitches and these de­lay pro­duc­tion. Sim­i­larly, as any­one who’s ever launched a new prod­uct will tell you, it’s very hard to re­li­ably pre­dict how long it will take to launch the prod­uct and how quickly cus­tomer de­mand will ramp up.

So when your ser vers need to be or­dered and paid for, then de­liv­ered and tested, you have a lot of work­ing cap­i­tal locked up in in­fra­struc­ture. This is not ter­ri­ble news if the prod­uct is de­liv­ered on time and things take off. How­ever, when it takes longer than ex­pected to go live and longer than ex­pected for cus­tomers to start buy­ing your prod­uct, you now have ex­pen­sive servers sit­ting idle, gob­bling up cap­i­tal that de­mands re­turns of ±40% while the as­sets it bought are de­pre­ci­at­ing at 33% a year. You don’t need to do the maths to un­der­stand the ug­li­ness of this pic­ture from a cap­i­tal per­spec­tive.


In the age of op­ti­mism (as the 1999/ 2000 pe­riod was) it seemed like ev­ery­thing would be on­line in a few months and a mas­sive land grab oc­curred as start-ups and in­vestors fought for f irst-mover ad­van­tage. So along comes a start-up that wants to sell high fash­ion on­line. Al­though we chose not to in­vest, they had a very clever sys­tem which in­cluded a vir­tual avatar that would guide you through the process of buy­ing, just as an in-store as­sis­tant would in the real world. Mind-blow­ing stuff 15 years ago. In­vestors swooned and the en­trepreneurs raised £ 125m ( yes, over R2bn) – for a pre-rev­enue start-up. An es­ti­mated $70m of this was for tech­ni­cal in­fra­struc­ture.


High fash­ion comes with other prob­lems though: lead times on the cloth­ing it sel f. Cloth­ing de­signs must be cho­sen, or­dered, paid for and de­liv­ered in a process that takes months be­fore items are ready for sale. Fash­ion­able items, by def i ni­tion, are per­ish­able – miss the mar­ket by a few months and you’re sell­ing what was fash­ion­able, not what is fash­ion­able. Or rather, you’re not sell­ing.

So add to­gether tech­ni­cal de­lays, in­evitable launch prob­lems, and a sys­tem that was de­signed for more band­width than most users had at the time, plus a very poor user in­ter­face and you end up with one of the big­gest dot-com busts of all time. launched late, couldn’t sell its no-longer-fash­ion­able cloth­ing and ul­ti­mately died within 18 months. On liq­ui­da­tion, $70m of hard­ware and soft­ware got sold for $250 000.

This wouldn’t hap­pen to­day, largely thanks to Ama­



Ama­zon ob­vi­ously uses huge amounts of com­puter pro­cess­ing power, band­width and stor­age to drive its e-com­merce busi­ness. The com­pany knows how fast the costs come down, have scale in pur­chas­ing these ser­vices, and also have ex­cess ca­pac­ity in their net­works. It fig­ured out that it could lease this ex­cess ca­pac­ity to the mar­ket, which in turn would give the on­line re­tailer more scale and al­low it to bring down prices.

So it launched Ama­zon Web Ser­vices (AWS) in 2006. On AWS, you can lease a server by the hour, stor­age by the gi­ga­byte per month, and band­width is su­percheap. This sim­ple con­cept changed the start-up land­scape for­ever. Whereas spent $70m on in­fra­struc­ture, mod­ern start-ups spend about $1 000 to $5 000 a month (and this at scale – you can run a lot – my us­age of the sys­tem rarely tops $300 a month). There is no work­ing cap­i­tal re­quire­ment – you can sim­ply f ire up a server (in­stalled from an ex­ist­ing disk im­age, so no test­ing re­quired), give it some t ra f f ic , and shut it down as t ra f f ic d ies down. Things can run idly if you like or you can lease large amounts of ca­pac­ity if you’re run­ning some com­puter sim­u­la­tions, for ex­am­ple. Sci­en­tific ex­per­i­ments that took months run­ning on a few ma­chines now take days run­ning on hun­dreds. Ama­zon changed the world of tech start-ups by re­mov­ing the need to spend pre­cious cap­i­tal on in­fra­struc­ture. You sim­ply lease it now. At huge scale it still makes sense to build and buy your own, but there are very few businesses that will ever reach these lev­els.


The end re­sult is ref lected in the fund­ing needs of to­day’s start-ups. Seed cap­i­tal of $100 000 will achieve as much as $2m did 15 years ago. Add that to very cheap mar­ket­ing and cus­tomer ac­qui­si­tion po­ten­tial thanks to Google Ad­Words, and you have the abil­ity to build an ap­pli­ca­tion, build de­mand, and pro­vide a busi­ness model for un­der R1m.

If things are work­ing and you need to scale up then you have all the num­bers to con­vince in­vestors, who should still have enough cap­i­tal to as­sist.

Newspapers in English

Newspapers from South Africa

© PressReader. All rights reserved.