Yeah it's a good comparison to make, I also agree with the...

Yeah it's a good comparison to make, I also agree with the sentiment re:healthy discussion...

We all know companies are generally worth multiples of their revenue, and if SIP business is IP licencing then that market size value kind of gives you a value of the IP contained within it. In this case the SIP market annual revenue is worth $8bn the actual value of the businesses in that market we might assume are 20x the revenue, so... the market is could be worth $160bn if you tried to buy up all the companies / IP in the SIP market.

4DS are taking this to a point where where the IP is proven in industry relevant processes (that is what PL5&6 do). So at that point the IP is fully cooked and IF going down a licencing path the value of the company could then be worth 10-20x the revenue it could generate in a licencing arrangement.

if DM thinks it's a $1bn company then that could mean as a licencing business it could expect $50m to $100m/yr revenue (which is 1/10 to 1/20 of the $1bn DM estimate).

i think a reasonable approach (even if the assumptions and numbers are wrong) is to figure out the size of market it would play in, then make assumptions about the slice of the market it would take and then an assumption about the value a company would pay for the access to IP that allows them to participate in that market.

For example the "embedded memory" market is annually $750m. No matter how favourable the assumptions, can't get 4DS to a $1bn valuation using that market.

As a counter example, the high bandwidth memory (HBM) market, a subset of DRAM is an interesting one. HBM appears to be the type of memory used onboard NVIDIAs B100 GPUs used in AI (each of those units sells for 40k a pop). SK Hynix, Micron and Intel have all been memory suppliers to NVIDIA for their previous iterations of this GPU.

HBM is apparently 20% of the $110bn DRAM market, so close to $22bn annually.

if 4DS gets it right, a new architecture is required for how the memory would be best used in AI applications... By the time that's all worked out and with the expected growth in the market, we could make an assumption the new product category 4DS is in could be anywhere between $5-20bn (i.e. a share of the existing HBM market or something that pops up alongside it or replaces it).

assume a memory manufacturer would pay say 10% of their revenue to access the IP that's $500m-$2bn.

Assume value:revenue of say 5:1 (5:1 represents the discount you'd apply today for it not being developed and market ready) - that's a $2.5 to $10bn value.

all of that is assuming that usual DRAM and Flash used in AI architecture is retained (E.g. DC applications for DRAM and flash on top of the GPU). If all three can be replaced by 4DS tech then you're talking about a slice of all three of those markets. Which is massive.

of course this is all speculation and you can change any of those assumptions, it moves up and down pretty quickly.


So when I do this, I can see where the optimism comes from with the upside potential. I can see how there are more conservative estimates using different approaches.

i can also see where the hesitancy comes from based on today's share price...

hopefully the answer is in the middle of all of those OR it blows our socks off. If it does, we can all afford to buy new socks.