I made my first Breakthrough Tech Profits recommendation last week, and I need to take just a second to let you know how excited I am about joining Jim Pearce and the rest of the great BTP team in finding winning tech stocks.

Today’s recommendation is a must-own for any growth-focused tech portfolio. It also reflects my approach to investing.

I work to keep abreast of the big-picture strategic trends as well as shorter-term market currents. I’m a regular reader of securities filings. But I don’t care to guess next year’s revenue to the nearest million and don’t like using earnings multiples to try to discern value.

I think people who do mostly use such guesstimates to lull themselves into a false sense of security. Investing is a highly uncertain undertaking: if it weren’t everyone would do it well and we’d all be on the beach attended to by our robotic servants by now.

Which might still happen one day. As artificial intelligence with capacity to learn demolishes the best human players at Go and poker – games deemed until recently too complex for AI to crack – it becomes harder and harder to figure out what tasks won’t be performed in the future by intelligent robots.

But learning, thinking robots can’t use a glorified laptop for a brain. Laptops and other computers take their marching orders from the CPU, or central processing unit, which can be further subdivided Into a few multiple cores. These are great at logic sequences but less efficient at the brute-force math needed to render graphics, simulate reality or stimulate AI learning and pattern recognition.

The human brain doesn’t have the equivalent of a CPU; instead its processing gets done in parallel by neural networks, each made up of thousands of neurons. That’s also how a GPU, or graphics processing unit, handles the extra math required to, for example, deliver hyper-realistic videogame visuals. It doles out the work to thousands of autonomous cores that work in parallel just like neural networks, sharing their calculations to assemble the full picture.

Source: Nvidia

Nvidia (NASDAQ: NVDA) is the undisputed champ of the GPU at the moment, dominating the high-end gaming market that still accounts for a bit more than  half its revenue. Demand for GPUs should keep soaring as virtual reality becomes more commonplace, since immersive VR requires seven times more graphics processing power than a PC videogame.

But that’s not why NVDA was the best-performing S&P 500 stock last year, more than tripling its price. Massively parallel processing isn’t just useful for obtaining the sharpest videogame and VR headset graphics. It also turns out to be essential for unleashing “deep learning” AI inside the data center and translating the surroundings of a self-driving auto into data the software can “see” to steer and brake. These world-changing technologies are in their infancy and Nvidia’s processors are their mother’s milk.

The reigning champion of central processing, Intel (NASDAQ: INTC), is sufficiently worried to have spent $30 billion on acquisitions expanding its reach beyond the CPU over the last couple of years. It would have been much better off buying Nvidia before its price soared; it may still have to resort to that.

So why are so many traders out there chomping at the bit to short this winner, and why have several sell-side analysts rushed to downgrade it in recent months? In short, human nature. Our innate anchoring bias informs us that something that has doubled in price in a few months can’t possibly sustain that, because “what goes up must come down.”

But that old saw often doesn’t cut it for stocks making big moves in either direction as the market prices in their altered situation. Nvidia briefly silenced the haters with blowout fourth-quarter earnings backed by 55% revenue growth year-over-year. And while the share price subsequently pulled back 18% from an all-time high in the course of a month through early March, it has since rebounded.

Nvidia has lots of competitors but no scale rivals in the market for GPUs, where growing adoption of its software tools for programming such chips gives it a considerable first-mover advantage and competitive moat.

That doesn’t mean it can’t suffer a steep correction somewhere along the line in the cyclical semiconductor industry, of course. It does mean that such a correction would present another buying opportunity.

“We’ve been investing in a lot of startups applying deep learning to many areas, and every single one effectively comes in building on Nvidia’s platform,” leading venture capitalist Marc Andreessen recently told Forbes magazine. “It’s like when people were all building on Windows in the ’90s or all building on the iPhone in the late 2000s. For fun, our firm has an internal game of what public companies we’d invest in if we were a hedge fund. We’d put all our money into Nvidia.”

You don’t want to do that. But you also don’t want to not invest in it based on irrelevant facts about how mispriced it was at the outset of 2016, or what its price to trailing sales looks like.

Nvidia is pricey, yes – most stocks are these days. But it’s also delivering growth like almost no one else, and will ultimately be valued not on last year’s sales or next year’s earnings but on the larger and longer-term opportunity.

The company is sitting on $4 billion in cash net of debt and spending $1.25 billion this year on share buybacks and dividends. Sales are on a trajectory to double within three years, and come with gross margins of nearly 60% that are likely to prove resilient.

We’re adding NVDA to the portfolio with a buy limit of $120.