Earlier this year, Rambus CEO Ron Black gave a keynote presentation at the IESA (Indian Electronic and Semiconductor Association) Vision Summit in Bangalore. Black’s presentation was covered by a number of prominent publications, including Rajeev Srinivasan of SwarajyaMag.
As Black told attendees, a new semiconductor paradigm may evolve in the aftermath of the recent $200 billion M&A activity which resulted in massive industry consolidation.
“He suggested that the selling price of devices (though obviously not the cost) will tend towards zero, i.e. you will have to give the devices away, and make your money through affiliated services,” Srinivasan reported. “There is precedent for what he says about shifts in where the value lies—there have been several clear phase shifts in the past.”
* 1950s & 60s Mainframe computers: IBM
* 1970s Minicomputers: DEC/DG
* 1980s Engineering workstations: Sun/Hewlett-Packard
* 1980s & 90s PCs: Microsoft/Intel
* 1990s Internet: Cisco/Ethernet/TCP-IP
* 2000s Mobile: Nokia/Apple/GSM
* 2010s Cloud and IoT: Google/Facebook/Amazon
“In the first three phases, the hardware was king, and you used to give away the software just to sell boxes. In the fourth phase, Microsoft figured out how to shift the value from the hardware (commoditized) to software (Windows),” Srinivasan explained. “In the sixth phase, hardware came into its own again. But in the current phase, the value has once again shifted, from the platform to the data. The data owners now capture the value, and they are willing to give away both platforms and software to capture data. The era of mass commoditization of hardware has arrived.”
However, as Black points out, most hardware makers have little or no access to the data or generators of data, as this space is typically dominated by intermediary data services companies. In the future, says Black, hardware makers may seek to generate annuities from data services providers as a way of amortizing device manufacturing costs.
“Ron Black’s intriguing proposal is as follows: semiconductor makers should incorporate mechanisms that lead to service revenue for themselves post-deployment. As an example, he suggests a security mechanism in chips that could be a silicon ‘root-of-trust,’ using cryptographic keys,” Srinivasan elaborated. “Using this mechanism to ‘attest’ that a chip is authentic could be a first-order service, especially when there is an epidemic of fake chips. There are many instances where chips have been swapped in and used for nefarious purposes.”
As Srinivasan notes, Black also pointed out that a root-of-trust can be leveraged for additional services.
“Imagine a field programmable gate array (FPGA). A trusted service entity with the right key will be able to reprogram the FPGA while deployed in the field, turning on a CPU, turning off a GPU, and so on, as demand fluctuates: and that is another valuable service which may, for instance, increase the device’s life in the field by reducing its power consumption, or by matching peak computing power to peak demand,” he added.