Posts / SMCI / #POST-210731

What a UAE Deal Might Mean For SMCI

May 19, 2025 · 02:22

R/WALLSTREETBETS

↗ https://reddit.com/r/wallstreetbets/comments/1kq0ugj/what_a_uae_deal_might_mean_for_smci/

This has been a pretty eventful week for SMCI and the AI sector as a whole. the acquisition of a [$20 billion deal](https://www.datacenterdynamics.com/en/news/saudis-datavolt-signs-20bn-deal-with-supermicro/) with DataVolt in Saudi Arabia and a 39% jump in share price off the strength of the news and overall AI sentiment improving is enough to make any degen hard as solid veiny diamonds. One thing this week that has especially caught my eye is UAE's plan to construct a [5GW AI Campus](https://www.commerce.gov/news/press-releases/2025/05/uae-and-us-presidents-attend-unveiling-phase-1-new-5gw-ai-campus-abu), with the help of G42, that will house 5GW worth of data center racks and span **10 square miles**.

To put this into perspective, thats:

* 50k 100kW full-scale liquid-cooled racks
* 3.2 **million** Nvidia Blackwell (B100) chips
* More power than NYC and San Francisco **combined**
* Equivalent to powering **3-4 million** homes or running **50 million** gaming PCs at full power 24/7

As you can see, this is an insanely ambitious feat the UAE is trying to achieve. Nonetheless, I couldn't help but ask myself, of the roughly 50k racks the UAE plans to procure, how many of those racks will be SMCI? Furthermore, whatever amount of racks SMCI does provide, how much money does SMCI stand to make from the project? Most importantly, What does that mean for the share price?

First, we need to understand what DCBBS is, and how it actually benefits customers with large scale projects like the UAE AI Campus.

DCBBS stands for **Direct Connect Backplane Building Block Solutions**, and it’s SMCI's proprietary system architecture for building **full AI racks,** not just individual servers. This means when a customer selects which AI rack configuration they want, when it arrives at their door, it comes in a form [like this](https://trendspider.com/blog/dell-and-super-micro-power-musks-ai-ambitions/). In other words, customers are getting Kitchen-Aid refrigerators sent to their fucking door that contain 30-90 Nvidia Blackwell chips and liquid cooling systems within them. No more receiving your SMCI rack order in individual server pieces that you have to assemble on your own. This greatly increases the speed in which these new data centers can go live. In most cases, these racks can be installed and plugged in in **under a day.**

Additionally, this also helps to increase their margins over time.

Traditionally, SMCI sold servers as modular building blocks. things like:

* 1U and 2U server chassis
* Motherboards
* Power supplies
* GPU trays

Customers (like system integrators, VARs, or hyperscalers) would assemble these into racks themselves. That’s low-margin business, because you’re selling parts, not solutions. the DCBBS racks contain:

* CPU & GPU servers (e.g., Nvidia Blackwell, AMD EPYC, Intel Gaudi)
* Total liquid cooling (cold plates, manifolds, chillers, etc.)
* Network switches
* Rack management software
* Power distribution units (PDUs)

Why does this new model increase margins? Because it is a shift SMCI is making, from selling modular pieces that must be assembled, to selling complete solutions that are pre-assembled and ready for use upon arrival. This gives SMCI increased pricing power and in turn, increased margins. Below is a comparison table to illustrate the logic:

Now lets go more in depth:

**1. Higher Average Selling Price (ASP) per Unit**

* Traditional air-cooled servers sell for **$10K–$50K** each.
* Liquid-cooled **DCBBS racks** (often 64–96 GPUs) sell for **$2–5 million per rack**.
* Even if gross margins stay in the 8–10% range, **gross profit per rack is 40×–80× higher** than per-server sales.

Higher ASP = greater dollar-margin per unit sold, allowing SMCI to increase overall margins even with a similar percentage.

**2. Bundled Value = Control Over Margin Levers**

With DCBBS, SMCI is not just selling a server, it’s delivering:

* 8–12 liquid-cooled GPU servers
* 2000+ liquid-cooled racks a month
* Integrated CDUs (coolant distribution units)
* Power delivery + networking
* Cabling, plumbing, and chassis integration
* Software/BIOS optimization

All of these are designed, manufactured, and **assembled in-house**, which means:

1. Vertical integration reduces dependency on low-margin commodity components
2. Custom bundles let SMCI charge more for value-added services
3. Standardization improves yield, reduces warranty risk, and shortens labor cycles

**3. Faster Time-to-Deployment = Premium Pricing Power**

SMCI can ship **fully integrated, plug-and-play racks in 3–4 weeks**, while others take 8–12 weeks+ with multiple vendors.

For hyperscalers and AI startups racing to bring GPU clusters online, this **speed translates into willingness to pay more**, or at least not negotiate down.

For example: A customer buying a $2.5M rack from SMCI may avoid:

* Extra $500k/month in AI cloud rental costs
* 3–4 month delay in deploying AI training models
* Facility delays in retrofitting for traditional HVAC-based servers

Below is a table below comparing liquid-cooled racks Time-to-Deployment between SMCI and its competitors:

This “time-value” premium enhances margin per contract.

**4. Liquid Cooling = Lower Total Cost of Ownership (TCO) for Clients**

Clients benefit from:

* 30–50%+ reduction in power usage (no fans/blowers at server level)
* Higher rack density → fewer racks → less real estate
* Reduced HVAC burden → capex + opex savings

Because DCBBS racks are pre-validated and ready for 80–100+ kW loads, customers can safely go denser per rack — which:

* Makes SMCI’s per-rack offering more economically valuable
* Lets them preserve pricing vs. air-cooled OEMs like Lenovo, HPE, or Dell, who may cut prices to remain competitive

Higher value-to-customer means more pricing power = higher gross margin

**5. Operational Efficiency at Scale**

SMCI is building out:

* A 3 million square foot manufacturing campus in Silicon Valley and Malaysia
* A supply chain optimized for GPU pre-allocation, liquid cooling parts, and just-in-time rack assembly

This allows SMCI to:

* Mass produce racks with minimal retooling
* Reduce per-rack cost at high volume
* Use contract-based repeat orders (like the DataVolt $20B deal) to plan capacity

Economies of scale + predictable assembly processes = declining cost/unit → rising margin

Below is a comparison table illustrating the margin impact:

So far, we've learned how DCBBS can substantially decrease build times for customers and how it could potentially increase margins for SMCI.

Now to the fun part: The UAE Deal

After doing some ape fisted carpet eating math, I estimate that the total cost of the 5GW AI Campus would be about $125 billion (assuming **each rack** draws about 100kW and sells for $2.5 million). Based on these numbers, I had ChatGPT build a table with a break down of what SMCI would get based on various percentages of the project they would be responsible for and superimposed a likelihood of occurrence formula in there:

|Coverage %|Likelihood of Occurrence|Racks Provided|G42 Total Payment to SMCI|Estimated Gross Profit (USD) 9.7%|Estimated EBITDA (USD)|Estimated EPS|Implied Share Price (@20x PE)|
|:-|:-|:-|:-|:-|:-|:-|:-|
|1%|Highly Likely (90%)|500|$1,250,000,000|$121,250,000|$103,062,500|$1.87|$37.40|
|2%|Very Likely (80%)|1000|$2,500,000,000|$242,500,000|$206,125,000|$3.75|$75.00|
|5%|Likely (65%)|2500|$6,250,000,000|$606,250,000|$515,312,500|$9.37|$187.40|
|10%|Possible (45%)|5000|$12,500,000,000|$1,212,500,000|$1,030,625,000|$18.74|$374.80|
|25%|Plausible (25%)|12500|$31,250,000,000|$3,031,250,000|$2,576,562,500|$46.84|$936.80|
|50%|Unlikely (10%)|25000|$62,500,000,000|$6,062,500,000|$5,153,125,000|$93.70|$1,874.00|
|100%|Very Unlikely (3%)|50000|$125,000,000,000|$12,125,000,000|$10,306,250,000|$187.40|$3,748.00|

As you can see, even with a conservative PE of 20x and 9.7% margins, if SMCI gets even 2-5% of the project, thats an EPS of $3.75-9.37 and a share price of $75-187 a share.

**How ChatGPT Calculated the Likelihoods**

This wasn’t a statistical model but rather a strategic probability estimate based on 5 weighted factors:

**1. Historical Precedent (30%)**

* No hyperscaler or government has ever given 100% of a 5GW AI deployment to one vendor.
* Datavolt gave SMCI a $20B deal, but likely part of a larger multi-phase build — suggesting up to 25% is plausible.

**2. Industry Norm for Diversification (25%)**

* Large clients (e.g. AWS, G42, Meta) **split procurement across 2–4 OEMs** for risk mitigation, price leverage, and geopolitical hedging.
* Most national-scale AI campuses split rack-scale work among **Dell, Lenovo, HPE, SMCI, Nvidia integrators**, etc.

**3. SMCI Production Capacity (20%)**

* SMCI can output \~5,000 racks/month total, **2,000+ liquid cooled**.
* To hit 50–100% coverage of a 5GW facility, they’d need to commit 6–12 months of global capacity, which limits probability unless it's multi-year.

**4. Geopolitical Alignment (15%)**

* SMCI already has deep regional partnerships (Saudi Arabia, Abu Dhabi), so 10–25% is well within reach.
* But **50–100%** would likely require **exclusive national alignment**, which is rare unless the vendor is sovereign or majority-owned.

**5. First-Mover Advantage & Time-to-Deploy (10%)**

* SMCI’s lead time (3–4 weeks vs. 8–12 weeks) is a massive advantage.
* For urgency-sensitive contracts, this could **justify awarding 25% or more** — but not 100%, barring a disruption in Dell/HPE supply.

I think its also important to add, two job listenings have been posted for positions in Abu Dhabi:

[Account Sales Manager](https://jobs.supermicro.com/job/United-Arab-Emirates-Abu-Dha-Sales-Account-Manager/1148026500/?utm_source=chatgpt.com)

[Technology Enablement Engineer](https://jobs.supermicro.com/job/United-Arab-Emirates-Abu-Dha-Technology-Enablement-Engineer/1246502400/?utm_source=chatgpt.com)

All in all, SMCI could stand to make a assload of money from a G42 UAE deal.

Last but not least, Cramer is [“Sick of”](https://www.msn.com/en-us/money/topstocks/jim-cramer-is-sick-of-super-micro-computer-stock-should-you-sell-smci-now/ar-AA1F0NkN?ocid=finance-verthp-feeds) SMCI

TL:DR: SMCI provides the fastest Time-to-Deployment of any of its competitors with its liquid-cooled DCBBS solution and this could help increase margins overtime and make it a more attractive buy to larger customers ready to deploy immediately. Capturing even 5% of the G42 UAE deal could result in a share price of $180+ a share. SMCI is already hiring for positions in Abu Dhabi.

Position: 15 $91c 1/16
https://ibb.co/b5J9VNmY