In this lecture we draw on material from Chris Anderson's book "Free" to explore how free products have changed over the last 20 years. The exponentially falling costs of CPU, storage and bandwidth mean that the low marginal costs of serving today's customers will continue to halve every year allowing smart companies to round down into the future.
2. Overview
Cost reduction in CPU, Storage and Bandwidth
Why Moore’s Law works
Atoms vs. Bits
Inflection point in exponential growth
Impact of Cloud computing
3. Triple play of
technologies
Transistors (CPU)
- Part of the computer that does the calculations
Storage
- Computers memory
Bandwidth
- Network that allows computers to exchange
information
All three are falling in price exponentially
4. Transistors
In 1961 a single transistor cost
$10
- 2 years later it was $5
- By 1965 it was $2.50
- 1968 $1
- 1975 $0.10
- 1982 $0.01
Today
- Intel 62-Core Xeon Phi: 5B
transistors
- $2500 -> 0.00005 cents
- 16Gb flash drive 64B
- $10 -> 0.00000002 cents
http://en.wikipedia.org/wiki/Transistor_count
Gordon Moore
5. Rounding down
Early 60’s Fairchild Semiconductor 1211: $100
Competing with vacuum tubes: $1.05
Robert Noyce, Jerry Sanders rounded down
- Knew that volume production would lower price
- Offered transistors for $1.05 from the start
“We were going to make chips in a factory we hadn’t built, using a
process we hadn’t developed but the bottom line was that we
were out there next week quoting $1.05. We were selling into the
future.”
6. Hard drives
Flips magnetic alignment on a
spiral of spinning disk
Head floats 10s of nm over
the platter
Platter spins at 4,000-
15,000rpm
Performance improvements
from:
- Faster speeds
- Higher power magnetic heads
- Closer to the disk
- Higher quality magnetic
surface
Spinning disks replaced with
Solid State Disks
- No moving parts
Kryder's Law – price
performance of storage
doubles every 18 months
7. Bandwidth
Speed of optical switching
Ultimate speed limited
- Speed of light
Can squeeze more colors down
a single fiber
Building fast switches that work
purely with light
Doubling in price performance
every 9 months – 1 year
8. Why Moore’s Law Works
Learning curve
- One person gets better with practice
- 10-15% for each doubling in production quantity
Experience curve
- Many people get better together
- 25% for each doubling in production quantity
Technology innovation
- 1970’s saw 12 doublings in transistor production
- Experience curve predicts 1/13
- Actually observed: 1/1000 (!)
- Ideas based businesses
- Sustained improvement through continuous innovation
9. Atom vs. Bit Economy
Rules are different for atoms
New Free is about bits not atoms
- Falling cost of storing, processing and delivering bits
- Bits can be copied perfectly almost for free
- Near zero marginal cost
Price of digital goods is will fall towards to zero
10. 2nd Half of the Chessboard
Grains of wheat on a chessboard
Double the grains for each square
- (1 + 263)grains
Lesson in the unexpected nature of exponential growth
First half of the chessboard numbers are small
Second half (inflection point) numbers go crazy
CPU, Storage and Bandwidth are entering the 2nd half of the
chessboard
No indication that they are slowing down
11. Reach for the clouds
Scale important to reduce fixed costs
8 years ago had to buy servers
The Cloud changed the equation
- Start-ups can take advantage of
economies of incremental server
costs
- Web start-up needs only a credit card
- Costs scale with service
- Not as cheap as google but
- when users numbers are small costs
remain manageable
- New entrants to free have an
advantage
12. Summary
Cost reduction in CPU, Storage and Bandwidth
Why Moore’s Law works
Atoms vs. Bits
Inflection point in exponential growth
Impact of Cloud computing