Description

AI is a fascinating trend and one that no doubt will continue to dominate news. 

Globally AI has bloomed like a rose,
Vital Components, the market chose.
But the bubble inflated,
Soon may be investors, frustrated,
Might find their dreams popping in prose!

Hat-tip to Chat GPT for coming up with this little ditty to help set the table.  Now here are some quick background facts:

Asia Vital Components (“AVC”) is a Taiwanese based manufacturer of fan coolers for CPUs, laptops, PCs, servers, etc.  They also provide thermal products and peripherals including heat sink and thermal modules. 

AVC is +175% YTD and +621% from pre-COVID.  From IPO through YE 2019 the TSR (with dividends reinvested) on the stock was 6.27%.  From YE 2019 through today the TSR (with dividends reinvested) on the stock was +74.3%.  COVID + AI has been good to folks who make PC/server fans.

AVC originally IPO’d in late 2002.  Using 2003-2022 financials we see the following:

Average Revenue CAGR: +10.8% (pretty good IMO)

Average Gross Margin: +14.2%

Average EBIT Margin: +4.32%

Average FCF Margin: +1.67%

Here is where 2023 is expected to come in:

Gross Margin: +20.8% (avg is +14.2%)

EBIT Margin: +12.4% (avg is +4.3%)
FCF Margin: +8.4% (avg is +1.7%)

There are three main divisions to the business:

(1)   Thermal

(2)   Chassis

(3)   Fositek + Peripheral

The following chart shows segment gross margins over time.  Bulls will have you believe this is an AI story, etc.  but its simply just an ASP story (historically negative ASP business that shifted positive in COVID then happened to catch lightening in a bottle with some AI stuff that will elongate it but will attract competition leading to a margin reset):

As can be seen above, every segment is substantially above its LT average and this starts right around COVID when everything ticks up at once.  As supply capacity is added in thermal we should see gross margins revert and the overall mix go lower.

AI is obviously the big sexy and resides in the thermal (have to cool all that robot thinking).  AI gross margins are in the 25%-30% range for AVC vs 20%-25% for the traditional business (some bulls are suggesting AI margins could be as high as 40% but this is not sustainable given low barriers to entry – see more on this below).  AVC management already acknowledges there could be ASP and gross margin pressure over time. 

For FY23E AVC will only derive 4% of its revenue from AI related business.  For FY24E (using sell-side consensus) this is estimated to rise up to a towering 7%.  The AI portion of the profit mix for FY23E is 8% and for FY24E this is 11%.  Point being unlike other hardware businesses with vastly greater AI exposure this is not that. This is literally small fans and water cooling pipes – not cutting edge anything…

AVC will not ship into the B100 with their liquid cooling (3DVC for A100 was a 50/50 split between Cooler Master and AVC).  3DVC = 3D vapor chamber. For the H100 it is sole sourced by Cooler Master.   The H100 SXM’s power consumption is 700W so it requires 3DVC. As for mass production of H100 SXM, Cooler Master serves as Nvidia’s primary supplier while Nvidia also has Furukawa as its secondary supplier. AVC has a yield rate that is 40-50% lower than Cooler Master’s.  The major challenge with 3DVC lies in the capillary structure located at the welding position between heat pipes and VCs. While the material costs of melding vertical heat pipes in 3DVC may be similar to that of using eight heat pipes to weld an ordinary VC, quotations for 3DVC can be doubled due to the lower yield rate and production processes involved.

AVC is hoping to take advantage of this share starting in 2024 as they open up capacity in Taiwan.  Right now there is 24-32 fans per rack and gross margins for AI fans are in the $30 ASP with 25%-30% range.  Many bulls on the street think the ASPs can go higher but with lots of competition and a non-industry leading position (Cooler Master is the leader) it is hard to push pricing power. 

Right now AVC primarily sells liquid cooling to Chinese customers.  They do not sell liquid cooling to SMCI who has its own supplier.  AVC competes with: Auras, Delta, and Cooler Master. 

Overall this is a business with significant competition – frankly its why this business has averaged fairly mediocre gross margins for so long.  For the world to suddenly think this has magical pricing power and cooling has gotten insanely more complex is really missing the boat. 

In the regular server business AVC has been gaining share.  They expanded from MSFT/AMZ (their initial customers) and have added some of the other Magnificent Seven (GOOG + META).  The main reason gross margins expanded during covid was a business mix shift – the chassis business used to be 80% PC chassis (which was 0% or loss making) and today its 80% server which is profitable.  So its somewhat unfair to compare to LT historical averages – but as shown I don’t think it will get back to those lows….just lower from today’s highs.  At present AVC has significant inventory sitting at distribution hubs which are waiting for customers to clear – being a Taiwanese company you get monthly data on this and the October monthly data was weak as H100 has not ramped yet so they have a bunch of products just sitting out there.

This has all the hallmarks of a traditional constrained capacity supply short (worth flagging this industry has historically declining ASPs in between product cycle launches which is why on average gross margin trends lower except for new product spikes….we are in one of those now).  Demand shot up making capacity constrained – gross margins shot up with ASPs and now huge amounts of capacity are being added potentially making gross margins revert to some prior level (below the current peak).  In the past period we saw strong ASP increases driven by platform upgrades like Eagle Stream. But in other areas we have seen pricing declines for graphics cards and potential inventory control in that segment suggest non-server areas face weaker near-term pricing dynamics.  The same will be true in the server segment as supply comes online. 

Here are some announcements of recent capacity expansion:
AVC: https://www.digitimes.com/news/a20230809PD215/avc-cooling-module-demand-vietnam.html
Aura: https://www.digitimes.com/news/a20231115PD215/auras-technology-liquid-cooling-ai-nvidia.html
Delta: https://www.digitimes.com/news/a20230907PD203/china-delta-electronics-ev-ict-manufacturing.html

As capacity comes online in 2024 we will start to see more ASP declines (we can track this actively with monthly disclosures). 

How much can we make?

Bull argue with the new AI fanciness here and sustainably higher margins (ignoring competition + large oncoming supply) this should trade at 24x P/E vs the long-term average of ~11x (see below):

Using some finely-tuned monkey math.  If this earns something like TWD$75B in 2025E (street is at $80B) and does so at a 8% EBIT margin (LT historical average is 4.3% and they will do +12.4% this year) that will result in ~12 in EPS.  If they get their LT average P/E of 11x and we hold the dividend flat at the predicted new level of 4.60 TWD per year (1.55% yield) that implies a 40% downside and -20% IRR on the short.

As in all industries, especially tech hardware, this time is different stories are usually great ones to bet against.  With the already high “peak” this resides at given the YTD rally it’s a good spot to start.

Catalyst

Capacity additions + normalized ASPs as newly launched products age out => lower revenue + gross margins => lower EPS. 
AI can continue to be amazing and change the world - people just have to remember how low margin server fans + liquid cooling tends to be over time as its not fancy / complex stuff.