There has been a lot of discussion lately in the venture community about the favorability of direct listings over the IPO process. The chief complaint of IPOs is that securities are underpriced at issue. To better understand the comparison between the two methods, we took a look at 25 of the top tech IPOs in the last two years.

Before we look at the underpricing argument, it’s helpful to look at the cost of IPOs and direct listings. At the most basic level, the cost of a direct listing and traditional IPO are relatively similar in our sample. We found that the average company paid between $29.6M and $34M in underwriting fees based on the level of exercise. It’s important to note that this doesn’t include any fees incurred by sellers. Total underwriting fees in deals where sellers sold shares as part of the IPO process ranged slightly higher, an average between $32M to $37M. The two direct listings to date, Spotify and Slack, paid financial advisory fees of $35M and $22M respectively.

An appeal of the IPO process is that it allows companies to raise additional capital, which companies aren’t currently able to do during a direct listing. In our sample, the average company raised between $794M and $913M during the IPO process. This translates to ~$26M raised for every $1M in fees. If we look at the fees that Spotify and Slack paid for a direct listing, the companies could have raised an estimated $937M and $590M respectively.

This brings us to the main point of the debate: the underpricing of IPOs. When comparing the issue price to the opening price on the first day of trading, our sample found an average underpricing of 38%. Of the 25 companies we looked at, only two, Peloton and Uber, opened below their issue price. If we add the underwriting fees and underpricing costs together, the true cost of an IPO is between $333M to $382M with only 10% of the cost from underwriting fees. The vast majority of this amount is value transferred to those who are allocated underpriced shares at issue.

Shares of NVDA were down 1% after hours following the company’s October quarter report. It’s worth noting, in the past three months shares are up 37%. We remain long-term positive on the Nvidia story given the company’s 3-5 year opportunity in providing GPU’s for datacenters and machine learning applications. Beyond five years, we expect autonomous mobility applications will support continued 15% plus revenue growth. Our takeaways from the quarter:

• Return to growth: Nvidia’s business will return to mid 30% revenue growth in Jan-20, after being down by an average of 17% over the previous 3 quarters. The return to growth is largely due to channel clean up within the gaming segment along with strength in datacenter. The Jan-20 revenue growth pace should be sustainable given the company is entering a year of easier comps following the crypto business 2019 bust.
• Soft Guidance: Jan-20 revenue guidance was 2% below the Street, attributed to softness in the gaming segment mostly offset by strength in datacenter. Gross margin guidance for Jan-20 was better than expected at 64.5%, vs 64.1% in Oct-19. As a point of reference, these margins are higher than Intel at 59% and AMD at 43% in their most recent quarters. The company noted to expect gross margins to further expand in calendar 2020 due to Nvidia’s transition to 7nm manufacturing at TSMC.
• Gaming: (down 6% y/y, accounting for 55% of revenue): Gaming was better than expected driven by RTX sales. As mentioned, the gaming business is expected to soften in Jan-20 due to seasonality from OEMs including HP, Dell, and Lenovo. Specifically, demand that would normally have fallen into the January quarter was pulled in the October quarter. Importantly, the gaming segment demand over the combined October and January quarters is essentially in line with expectations.
• Datacenter: (up 7% y/y, accounting for 24% of revenue): The majority of the earnings call was focused on the datacenter opportunity. Demand for datacenter products is being driven by hyperscaling for training and inference AI algorithms. For example, natural language-based interfaces (Apple Siri, Google Voice, Amazon Alexa) are powered by cloud computing hardware optimized with Nvidia GPUs which reduce latency. We expect datacenter to be the largest revenue segment in the next 3-5 years.
• Auto: (down 6% y/y, accounting for 5% of revenue): Today about half of auto revenue is related to infotainment and half is related to autonomous driving hardware. The infotainment center business is declining slightly compared to the autonomous driving segment which is growing at lumpy rates. On the call, the company suggested autonomous driving segment growth slowed measurably given major car OEMs are years away for rolling out autonomy at scale. Aside from Tesla, Nvidia’s hardware remains imbedded in all major car OEMs autonomy solutions. We advise patience when considering the auto segment opportunity and expect in 5 plus years, it will become an increasingly important part of the story.

• Investing in Nvidia has been a rollercoaster. We believe revenue growth will reaccelerate over the next 12 months to 30% plus after being down an average of 17% over the past three quarters. The company is entering a year of easier comps following the crypto business boom (2018) and bust (2019).
• While crypto is just under 5% of revenue, unwinding channel inventory over the past year had a negative 15% impact on revenue. As of Oct-19, we believe the crypto channel overhang has been cleaned up.
• The return to growth is reasonable, given Nvidia is a key hardware arms supplier to the long-term growth trends of gaming, AI/datacenter, and autonomous mobility.
• This Thursday (Nov 14^th) Nvidia reports its October quarter. We expect positive results from the company based on upbeat macro commentary from Intel and AMD’s recent earnings reports, specifically about Dec-19 growth.