Google Is Building Its Own Chips. Nvidia Should Be Paying Attention.

$352.51. That is where Alphabet closed on July 13, and it is below every moving average that matters. The 20-day SMA is at $357.04. The 20-day EMA at $359.71. Price is underneath both, which is the technical definition of a stock the market has stopped buying. RSI at 43.9 is neutral-to-soft. MACD just flipped bullish with the histogram at +0.37 — a green shoot, but not a signal. ADX at 17.8 says there is no trend at all. Alphabet is drifting, and the market is waiting for something to break it out of the range.

That something is Q2 earnings, and the specific question is not about search revenue — that is growing again, and Harris Oakmark's Q2 investor letter flagged accelerating search revenue as the driver behind the stock's recent rally. The question is about TPUs (Yahoo Finance, July 14; Seeking Alpha, July 14).

Tensor Processing Units — Google's custom AI chips — are the most strategically important asset Alphabet owns that most investors cannot value. The thesis, laid out in a Seeking Alpha Q2 preview titled "The AWS Blueprint for TPUs," is straightforward: Amazon monetized cloud infrastructure by renting it to everyone. Google can do the same with silicon. If Alphabet opens TPU capacity to external customers through Google Cloud — and the margins on that capacity are better than reselling Nvidia GPUs — then Google Cloud's profitability inflects upward in a way the Street is not modeling. Cloud revenue growth has been strong. Cloud margin is the unlock.

The competitive read is where this gets interesting. A 24/7 Wall St analysis published July 14 reports that Google is "quietly building silicon that threatens to cut NVIDIA out of its own customers' data centers" — citing custom TPU performance that is approaching parity with Nvidia's H-class GPUs for inference workloads, at a fraction of the cost. Morgan Stanley noted the same day that a competitor chipmaker "remains a core AI winner and a close number two behind NVDA," which is the sell-side's way of acknowledging the AI silicon market is becoming a two-horse race without naming the second horse. The second horse is Google (24/7 Wall St, July 14).

The stock does not reflect this optionality. GOOGL at $352 is below its 20-day SMA and EMA, with ADX at 17.8 — the market is treating Alphabet as a search company with a cloud business, not as a silicon company with a search business. The MACD flip is the first technical signal that sentiment might be turning. If Q2 Cloud revenue beats and the margin trajectory is up, the TPU thesis gets priced in. If Cloud margins disappoint, the stock stays range-bound and the silicon story stays a story.

The levels are clean. Resistance at $357 — the 20-day SMA — is the first gate. Above that, $359.71 (the EMA) and then the Bollinger Band structure. Support is less defined because the trend is so weak, but a close below $340 would invalidate the bullish MACD cross and put the stock in a technical no-man's-land. ATR was not available from the feed, but volume has been light at 15 million shares — the kind of tape where an earnings beat can move the stock 5 percent on minimal resistance.

Alphabet's Q2 report will tell investors whether the TPU investment is showing up in Cloud margins or whether it is still a capex line without a revenue line. Google spent years building chips that only it could use. If it starts renting them to everyone else, Nvidia's pricing power — the thing the entire AI trade is built on — has a competitor. The stock at $352 is not pricing that in. The earnings print will.