Shares of Rambus are down more than 50 percent today after a jury trial on Wednesday found Micron and Hynix Semiconductor not guilty of price fixing. Rambus had sued the two manufacturers for over $12 billion, claiming that they formed “a secret and unlawful conspiracy to kill a revolutionary technology, make billions of dollars and hang onto power."
“We are disappointed with this verdict as we believe strongly in our case,” Rambus Chief Executive Harold Hughes said in a statement. “We do not agree with several rulings that affected how this case was presented to the jury and we are reviewing
our options for appeal."
Micron stock is up by some 20 percent as of this writing, with analysts claiming that the decision removed an "overhang" on the company's stock. The DRAM manufacturer still faces significant challenges within its core market; the rapidly falling price of memory (16GB of RAM is less than $80 on NewEgg) has undermined profitability and led other analysts to be more neutral on the company's prospects.
Plenty of Bad Faith To Go Around
Rambus has spent more than a decade suing nearly everyone involved in DRAM production or who uses DRAM in their products. After so long, it's difficult to quickly summarize the list of judgements, appeals, reversals, and fresh lawsuits. We've investigated the issue in the past and had a number of off-the-record conversations with people who worked in the DRAM industry back when Intel and Rambus were pushing RDRAM as the Next Big Thing. There was plenty of backroom dealing—on all sides of the table.
Did the DRAM manufacturers want to move to RDRAM? Absolutely not. Did they knowingly infringe on Rambus patents? Probably, yes. Did Rambus try to patent technologies and methods being discussed at JEDEC having promised up front not to do so? Yes.
128 MB of PC800 RDRAM was $800 in 1999
Ironically, the one company that appears to have acted in good faith (albeit with a hefty dollop of self-interest) is Intel; Santa Clara invested billions in helping DRAM manufacturers ramp RDRAM production. Intel had its own financial reasons for backing RDRAM, but it apparently dealt honestly with the parties involved regarding its own intentions. Rambus' accusations of price-fixing don't jive with what we know of the manufacturing situation at the time. Even the manufacturers who were 100 percent on board with the RDRAM conversion had major problems validating RDRAM modules. One of the most significant issues was that there was initially no way to test RDRAM IC's individually.
When a manufacturer like Kingston or Hynix builds RAM, they test each chip before soldering them to a PCB. This wasn't initially possible with RDRAM, which meant that RIMMs had to be fully assembled before they could be tested. The serial
nature of RDRAM made it impossible to determine which chip was responsible for failure—a RIMM either worked or it didn't. If it didn't, the only way to isolate the bad IC was to manually swap chips from previously tested modules. This made any attempt to "fix" faulty modules extremely time consuming, particularly since a non-functional RIMM didn't necessarily have just one
The availability of improved testing equipment eventually solved these problems, but RDRAM's initial sky-high prices were driven partly by the difficulty of manufacturing parts that could run at 800MHz in 1999
and partly by validation issues. There was no need to resort to price fixing to inflate the price of RDRAM. After ten years and this latest high-profile setback, one might hope Rambus would lose its taste for litigation, but that isn't yet the case.