Obviously most of the diagram shows the Good Things which activating AMPK does. There appear to be three core activating signals; changes in Ca2+ via calmodulin, increasing cyclic AMP and that change in AMP+ADP:ATP triggered by exercise, low glucose, hypoxia and severely toxic doses of metformin. Here they are emphasised in blue:
There is only one inhibitory input shown, I've marked it in red:
When you follow this input back you end up with the one factor which suppresses AMPK signalling.
We know that metformin inhibits mtG3Pdh at normal pharmacological concentrations. From the Protons perspective this blocks the generation of the ROS essential for insulin signaling. No insulin signalling, no inhibition of AMPK. Which nicely fits in with this paper:
Insulin inhibits AMPK activity and phosphorylates AMPK Ser485/491 through Akt in hepatocytes, myotubes and incubated rat skeletal muscle
Of course metformin inhibits complex I (which drops ATP and so activates AMPK) at concentrations which put you in to the ITU with lactic acidosis, around 200micromolar in plasma. Recall this?
It also activates AMPK via inhibition of AMP-deaminase using tissue culture exposure of 10mmolar. That's 10,000micromolar, which would probably put you in to the morgue rather than the ITU.
A quick reality check suggests that taking one 500mg metformin tablet an hour before a bike race might just help you win... Somehow I don't think blockade of complex I would do that! Freeing up fatty acids by dumping insulin signalling might just do the job.