Description

“The obscure we see eventually. The completely obvious, it seems, takes longer.”

- Edward R. Murrow

We believe NVIDIA Corporation (NVDA) is a short, poised for a significant decline in the share price (50%-75%) as several catalysts play out to re-rate the earnings trajectory and valuation over the next few months. The market has mistaken a major cyclical boom as a secular growth story, and the cyclical bust has already started in NVDA’s end-market as evidenced by excess inventory and dramatic price declines in NVDA’s products (discussed in further detail below).

A good writeup on VIC by afgtt2008, as well as various analyst reports, lay out the details on the business mix, so we don’t need to cover that ground here. Instead, we will lay out why the turning points for the NVDA story are increasing, and why the stock is likely to collapse on the coming earnings call (or prior, if current long holders and sellside analysts wake up to the problems).

We’ll address the thesis in four broad strokes –

1)      Why does the opportunity exist?

2)      What is our variant perception?

3)      How will the crowd come to our view?

4)      Fair value.

Why does the opportunity exist?

Over the past 3 years, NVIDIA has been the best performing stock in the S&P 500 index, having gained 1,143% since the beginning of 2015, far eclipsing the returns of Netflix or any of the other FANG or similar story stocks. It has been a momentum investor’s dream, and unlike many other momentum stocks that are just riding high on “story” without much tangible cash flow, NVDA has delivered accelerating growth in revenues, earnings and cash flows.

The growth trajectory has been explained by the Company as a variety of secular themes (artificial intelligence, driverless cars, etc.), and as a result, analysts have extrapolated the growth to continue. Earnings-based valuation metrics lead people to believe it is not “that” expensive at 42x trailing P/E or “only” 32x NTM P/E, a “reasonable” forecast assuming growth top line growth rates and margins based on recent history.

In addition, the stock is often promoted on CNBC by Jim Cramer, as well as pumped heavily into the retail channel through various banks’ private wealth management businesses (i.e. after recent Facebook “data scandal”, Bank of America Wealth Management replaced Facebook with NVIDIA on their “US Select 1” list for wealth management investors).

The investor base, and the narrative that NVIDIA management has created, combine to create a situation where the market is about to get blind-sided by a major cyclical bust in NVDA’s largest end market.

Our variant perception

Often the correct explanation for a situation is the most obvious one. We believe NVIDIA has undergone a major cyclical boom in demand for its core product (GPUs), due to a new source of demand (cryptocurrency mining, specifically Ethereum) entering an already tight supply/demand situation, rapidly pushing up prices during 2016-2018. As that source of demand wanes (Ethereum prices down ~67% since mid-Jan 2018), the supply/demand situation for GPUs broadly is shifting very unfavorably for NVIDIA, resulting in an oversupply of products and collapse in product prices, which has already started in recent months.

What does NVIDIA make? NVIDIA makes primarily GPUs (Graphic Processing Units), which are distinguished from CPUs (Central Processing Units) in computing applications. In computing, a CPU is thought of as the traditional “processor” within a computer that carries out the instructions of computer programs by performing basic arithmetic, logical, control and input/output (I/O) operations specified by the instructions. Over time, with increase in computer graphics, arose a need for specialized electronic circuits used primarily for graphics and image processing, and NVIDIA popularized the term “GPU” for Graphics Processing Unit in the 1990s.

Modern GPUs are used in a variety of applications for image processing, such as mobile phones, personal computers and video game consoles, among others. GPUs are more efficient at image processing, and their highly parallel structure makes them much more efficient than general-purpose CPUs for algorithms where the processing of large blocks of data is done in parallel (such a cryptocurrency “mining”).

NVIDIA management has tried to deflect risks from exposure to cryptocurrency mining, by stating that it’s a low percentage of revenues (in Fiscal Q1 earnings call on 5/10/2018, mgmt stated:

“Cryptocurrency demand was again stronger than expected, but we were able to fulfill most of it with crypto-specific GPUs, which are included in our OEM business at $289 million.” (this would be ~9% of revenues)

“Looking into Q2, we expect crypto-specific revenue to be about one-third of its Q1 level.” (implying cryptocurrency mining will drive ~3% of revenues in the forward quarter).

Investors have interpreted this generally as “cryptocurrency exposure is limited” and “baked into the guidance”. We disagree, and data from the GPU market is showing the flaw in this line of thinking.

We spent some time speaking with NVIDIA about how they sell into the cryptocurrency mining market, and realized how they are explaining the exposure to investors. The company goes out of its way to explain that they saw the opportunity in crypto mining, developed crypto-specific GPUs (without any graphics display functionality), and contract out directly to the large miners in Asia (Genesis or others). They focus the entire conversation on the direct sales to the cryptocurrency end market, and why it’s manageable.

We believe this is a major sleight-of-hand, as GPU chips made for NVDA’s core gaming market are interchangeable and able to be used for cryptocurrency mining, and in many of our conversations with large and small miners, cards through the GeForce gaming channel have been going heavily into the cryptocurrency mining end market. This has been evidenced by skyrocketing prices of GPUs, and top line growth rates and margins at NVDA in the past several years.

How will the crowd come to our view?

The turning point for NVIDIA’s business has already arrived, with the collapse of Ethereum prices in recent months, for which GPUs were heavily being used. Below is a price chart of Ethereum – after an impressive run up during 2017, where prices rallied from under $100 to over $1300 by January 2018, Ethereum prices have collapsed in recent months to sub-$500.

Various analysis of mining profitability indicate that at current levels, miners are facing poor returns on investment. In addition, the pending introduction of cheaper and more efficient ASICs for Ethereum mining (such as by BitMain) are likely to pressure GPU demand and prices for this end usage.

To paraphrase George Soros, we don’t attempt to “predict” how these impacts may play out, but simply “observe” what is happening in GPU markets in recent months.

We believe that GPU demand and profitability peaked in Fiscal Q1 (ending 4/29/18), and a large excess inventory situation and price decline of GPUs is underway, that is likely to intensify.

The chart below tracks the selling price of one of the top selling products, NVIDIA’s GeForce 1070. Since the peak of Ethereum prices in early January, product prices are down from almost $1000 to sub-$500, a decline of over 50%.

Even higher end products such as the GeForce 1080 are being heavily discounted across channels. Below is a promotion from reseller Newegg that is being run in the recent week on the GeForce 1080 for $499 (discounted from $599, but note that prices 6 months ago were over $1000).

As we were writing this up, awareness of this situation has started to percolate in the analyst community. Late last week, BlueFin Research Partners’ analyst Paul Peterson wrote after a two-week trip in Asia, as excerpted by Barron’s Online:

“GPU card sales dropped by more than 50% in the North America and Europe retail channels, according to our research”

“[not just cryptocurrency mining, but] gamer demand also appears relatively muted”

“[Nvidia was] surprised that gamer demand didn’t bounce back when GPU card prices plummeted”

“Nvidia is pressuring the retailers and add-in board makers to drop their prices in order to move the cards, leveraging future Turing availability in the balance."

Furthermore, Gigabyte Technology Co. out of Taiwan, a manufacturer and marketer of peripherals (approximately 50% of revenues from GPUs, other half from less volatile products), recently reported June revenues down 30% y/y. Gigabyte’s website has significant detailed historical data at https://www.gigabyte.com/Investor/81 and http://emops.twse.com.tw/server-java/t05st10_e?TYPEK=sii&step=historical&co_id=2376, from which we can see the trends so far in 2018.

Our takeaway is that GPUs are likely down over 40% year-over-year in June, which would also square with Peterson’s take, channel trackers of prices and heavy promotions we are seeing lately such as by Newegg.

At Gigabyte’s analyst meeting a few weeks ago, Chairman Yeh commented that crypto demand peaked in 1Q 2018, crypto was over 50% of graphic card volume mix in 1Q 2018, and that they have seen inventory build-up from slowing GPU sales since April.

Only two weeks ago, DigiTimes reports:

All of these speak to a significant deceleration (at best) in revenue growth at NVIDIA in the quarters ahead, let alone potential for significant YoY declines – at a time of expectations of 15-30% annual revenue growth for years to come by the investor community.

In addition, given NVIDIA’s historical EBITDA margins over the past several years, one can infer the likely trajectory of margins as GPU prices collapse:

2011: 16.4%

2012: 22.0%

2013: 20.9%

2014: 17.8%

2015: 20.9%

2016: 23.3%

2017: 31.0%

2018: 35.2% (Fiscal Year Ended 1/28/2018, which we believe was peak margins)

2019E: 42.1% (current street expectation for full Fiscal Year Ending 1/31/2019)

Fair value

It is difficult to accurately model the range of outcomes, and we don’t pretend to come up with false precision just for the sake of price targets. It is clear to us that NVIDIA is grossly overvalued and expectations are way too high compared to underlying reality, for both top-line growth and forward margins. If either are slightly missed, the downside is quite meaningful, creating a very asymmetric short opportunity.

To frame how asymmetric the outcomes are, a couple of model scenarios are contemplated. For the sake of multiple, we assume current framework (32x FY2019E EPS) is how the market will continue to treat the shares (generous assumption, given the top line and earnings slowdown we foresee).

Given the collapse in GPU prices, we believe it will be heroic for NVIDIA to generate 5% top-line growth this year, and reasonable scenarios point to -10% or worse (simply taking YoY changes in GPU prices, assuming units sold continue to increase significantly).

In addition, given that GPU prices have collapsed over 50% in various key products since January, it will be generous to assume only a 3-4% EBITDA margin hit; in a true realistic downside, margins going back to historical norm levels would imply a 15% EBITDA margin hit (in a draconian downside, there would be a period of even lower margins as inventory is cleared in the downcycle).

We will leave the reader to their own conclusions about which scenarios to run for sensitivities, but a rough framework identifies the range of possible downside outcomes below:

FY 2019E Rev Growth +5%, EBITDA Margin 35% = EPS $5.68 @ 32x = $176/share (30% downside)

FY 2019E Rev Growth +0%, EBITDA Margin 30% = EPS $4.54 @ 32x = $145/share (42% downside)

FY 2019E Rev Growth -5%, EBITDA Margin 25% = EPS $3.48 @ 32x = $111/share (55% downside)

We will leave it to the reader to contemplate whether the stock will maintain a 32x forward multiple if YoY earnings are flat or declining significantly in various of these scenarios. If the multiple declines, the downside is much greater, and it’s not hard to envision the stock down over 75% from current levels.

Other considerations

Trade tensions and increasing US-China tariffs may significantly impact NVDA’s margins going forward, as GPU firms have substantial cost exposure to Asian technology imports (given the declining demand situation, cost increases are highly unlikely to be passed through).

NVIDIA’s CEO Jensen Huang has been talking up pent-up gaming demand for months. However, at Computex recently, he said there’s likely no new GPU architecture to be rolled out by NVIDIA for quite some time. To us, this comment, coupled with NVIDIA’s significant discounting of products in the channel, indicates that management is surprised that price drops aren’t stoking gamer demand, and are very worried about the quarter.

We will leave the reader with some wisdom from Yogi Berra as it relates to NVIDIA’s stock:

“Nobody goes to that restaurant anymore – it’s too crowded.”

Catalyst

To paraphrase George Soros, we don’t attempt to “predict” how these impacts may play out, but simply “observe” what is happening in GPU markets in recent months. We believe that GPU demand and profitability peaked in Fiscal Q1 (ending 4/29/18), and a large excess inventory situation and price decline of GPUs is underway, that is likely to intensify. GPU prices are collapsing in various channels as discussed in depth in the writeup and data shown. NVIDIA's stock is likely to follow suit.