The Supply Chain Risk: Are we ignoring the next ASIC bottleneck?

[HashRate Hero]

Everyone is chasing hashrate efficiency/joule thinking that the secret is only technical ! , but we’re building a huge ilusion and huge non understanding of all the aspects that are controlling the operation.

At what point does it fail + Are we highly considering the  whole network aspect ?

I haven't seen much discussion on real resilience here are we just ignoring the hardware fragility risk?

[philipma1957]

Everyone is chasing hashrate efficiency/joule thinking that the secret is only technical ! , but we’re building a huge ilusion and huge non understanding of all the aspects that are controlling the operation.

At what point does it fail + Are we highly considering the  whole network aspect ?

I haven't seen much discussion on real resilience here are we just ignoring the hardware fragility risk?

Don't understand what you mean.

Mining is all price driven.

Price of power
Price of coins

completely override gear cost.

Ie if coins were  200k today gear price would moonshot.
If you have free or 1 cent or 2 cent power almost any gear make money.

S9's burning 1 kwatt do 12th at 1 cent power cost to run a s9 is 24 cents.

12th x 2.9 cents = 34.8 cents a 10 cent profit.

but if price is 126k they earn 69.6 cents they burn 24 cents and profit is 45.6 cents

and if price is 200k the s9 earns about $1.02 they burn 12 cents profit is 90 cents


so mining has a power cost issue due to AI and the lack of building out solar
and mining has an earning issue due to coin price as it has been flattened by EFT's

asic availability is a non issue right now.

and it will only be an issue if coin price moonshots and someone makes a fusion reactor thus making 1 cent per for all of us.

[logfiles]

Because almost everyone is mining so they can make something out of it... profits. So it's more than understandable that the primary focus will always be about efficiency. We will worry about the supply chain risks when they come. It's like you worrying about a pandemic that may never come by going to the supermarket and hording whatever groceries that you come across. 3–4 years later and nothing has happened.

[NotFuzzyWarm]

Any particular reason you pretty much made a dupe of this topic? https://bitcointalk.org/index.php?topic=5585737.msg66834501#msg66834501
That is not permitted as it splits up coherent discussion of your topic.

Now if you are referring to mining hardware supply - mainly chips - as I said in the other thread:

Two foundries make mining chips - TSMC & Samsung. If one of those 'goes down' there will far more industries affected than just mining as those companies make over 80% of all cutting/bleeding edge chips <5nm in the world. The only other two companies capable of <5nm chips are Intel and Global Foundries and they do not make mining ASIC'S. Think smartphones, video cards, cpu's, network çhips, etc. those are who would be affected with a far larger global impact.

Each company has several advanced node foundries so problems at one would be serious but not the end of the world. Considering the number of miners shutting down or shifting to chase the AI bubble there will be more than enough new or used high end mining gear to go around.

[HashRate Hero]

For flagging that supply chain risk, focusing solely on efficiency ignores the hardware's fragility.

It s normal to turn toward the need of starting to look past simple profit margins that says considering the bottleneck at the foundry level specifically the reliance on <7nm process nodes to truly assess our operational resilience.

That says and to  highlight that supply chain risk, focusing only on hashrate/joule ignores the foundry bottleneck. lets look at something like the BM1366 chip, any disruption at the <7nm node level would kill the production cycle for devices like the Bitaxe, away of how efficient the code is.

[NotFuzzyWarm]

And your solution would be what? Make chips using the larger nodes because there many more foundries that can make them?

Not going to happen because because the inefficiency of large nodes means that at current diff miners using those nodes will cost far more to run. Even with power costing 1¢ or less, diff would have to fall to what it was 5 or more years ago to make economic sense.

It is highly unlikely any new players will arrive making sub 5nm chips. TSMC's latest plant for 3nm chips being built in Phoenix AZ is costing them over $4 Billion and for at least the next 2 years most Engineers working there will be from TSMC's plants in Taiwan until US workers can be trained.

[HashRate Hero]

And your solution would be what? Make chips using the larger nodes because there many more foundries that can make them?

Not going to happen because because the inefficiency of large nodes means that at current diff miners using those nodes will cost far more to run. Even with power costing 1¢ or less, diff would have to fall to what is was 5 or more years ago to make economic sense.

Duplicated the post by error ; and i can't delete this one original post is in technical discussion  section ........... i will lock the other one .

To answer your last intervention i can say that efficiency gap makes older nodes a total non-starter margins just don't hold up.

My question to you is : Does mining just stop being about hardware specsto become nowdays a play for whoever can find the cheapest subsidized power?

[NotFuzzyWarm]

Does mining just stop being about hardware specs and strictly become a play for whoever can find the cheapest subsidized power?

Yes and it has always been that way. The size of a farm has always been a balance of cost to run vs profit. The size of mega farms like Foundry, MARA, et al has grown only because of more efficient miners while the farms cost per kWh for power stayed relatively fixed.

It was a no-brainer for them to not only replace older models with the latest/greatest ones but also add more to stay at their current power usage.

[HashRate Hero]

It was a no-brainer for them to not only replace older models with the latest/greatest ones but also add more to stay at their current power usage.

Exactly. It's always been a balance, but we're hitting a wall. Once supply chain premiums eat the gains from node shrinks, subsidized power is basically the only moat left

[NotFuzzyWarm]

It must be noted that chip node size is rapidly approaching the 1nm mark and there is no possibility of anything smaller because that is where any foreseeable photolithography processes end. Using direct ion or x-ray etching is far too slow to use for mass production processes.

As a Fun Fact, the size of an atom is between 0.1nm (hydrogen) to 0.35nm for uranium and sets an absolute minimum gate size.

[FP91G]

If improving the process technology doesn't significantly improve energy efficiency, then there's only one other option.
You can increase the chip size and improve cooling accordingly. If the new ASICs feature liquid cooling, this isn't such a big deal. I read that by 2029–2030 technology will reach the 1 nanometer mark, and that it will be impossible to reduce it further using classical methods.