NAND and NOR gates are universal logic gates: any Boolean expression can be built using only NAND gates, or only NOR gates, without any other gate type. Other gates such as AND and OR cannot do this on their own. Even so, designers usually prefer NAND gates over NOR gates. The main reason is physical, not logical: in CMOS, a NAND gate of the same speed is smaller and has less input capacitance than a NOR gate.
Why NAND wins: the short version
- Faster for the same size. In a CMOS NAND gate, the slow p-type transistors sit side by side (in parallel) and the fast n-type transistors are stacked in series. In a NOR gate it is the other way round: the slow p-type transistors are stacked. Stacking slow transistors hurts most, so a NOR gate has to make its p-type transistors much wider to keep up.
- Smaller and lighter on its inputs. Those wider transistors make a NOR gate larger and give it more input capacitance. That means it loads the gate that drives it more, which slows the whole path and uses more switching energy.
- The gap grows with inputs. A 4-input NOR is much worse than a 4-input NAND, so wide NOR gates are usually avoided altogether.
- History. In TTL logic, the basic gate is a NAND built around a transistor with several emitters, one per input. The 7400, a quad 2-input NAND, was the first part of the TTL family that became the industry standard.
What both gates can do
Because both are universal, NAND gates and NOR gates can each be used to build NOT, AND, OR, XOR and XNOR functions, and storage elements such as latches and flip-flops. The number of gates needed is symmetric: building OR from NANDs takes three gates, just as building AND from NORs does. So the preference is not about logic; it is about how well each gate performs in silicon.
A NAND gate also has a handy logical property: its output goes high if any input goes low. That makes it a natural detector for "at least one active-low signal is asserted", for example combining several active-low alarm or interrupt lines.
Deep dive: the numbers and details
This part holds the figures, equations and worked examples. Skip it if the overview is all you need.
How the transistors are arranged
| Gate | Pull-down network (nMOS) | Pull-up network (pMOS) | Transistors (2 inputs) |
|---|---|---|---|
| NAND2 | 2 in series | 2 in parallel | 4 |
| NOR2 | 2 in parallel | 2 in series | 4 |
| NAND n | n in series | n in parallel | 2n |
| NOR n | n in parallel | n in series | 2n |
Holes in p-type transistors move more slowly than electrons in n-type transistors, so a pMOS transistor needs to be wider than an nMOS transistor to carry the same current. The logical effort method commonly assumes a pMOS-to-nMOS width ratio (gamma) of 2 for a balanced inverter.
Worked example: sizing for equal drive
Take a reference inverter with an nMOS of width 1 and a pMOS of width 2, so its input capacitance is 3 units. Size each gate so that its worst-case pull-up and pull-down match the inverter. Transistors in series must each be made wider by the number in the stack:
| Gate | nMOS width (each) | pMOS width (each) | Input capacitance per input | Total transistor width |
|---|---|---|---|---|
| Inverter | 1 | 2 | 3 | 3 |
| NAND2 | 2 | 2 | 4 | 8 |
| NOR2 | 1 | 4 | 5 | 10 |
| NAND4 | 4 | 2 | 6 | 24 |
| NOR4 | 1 | 8 | 9 | 36 |
The NOR2 uses 25% more transistor width than the NAND2; the NOR4 uses 50% more than the NAND4. Logical effort g is the input capacitance divided by the inverter's 3 units: NAND2 = 4/3, NOR2 = 5/3, NAND4 = 2, NOR4 = 3. These match the general formulas g = (n + 2)/3 for an n-input NAND and g = (2n + 1)/3 for an n-input NOR in the logical effort tables.
Worked example: delay with a fanout of 4
In the logical effort model, the normalized delay of a stage is d = g x h + p, where h is the electrical effort (load capacitance divided by input capacitance) and p is the parasitic delay, which equals the number of inputs for NAND and NOR gates. For a fanout of 4 (h = 4):
NAND2: d = (4/3) x 4 + 2 = 7.33 NOR2: d = (5/3) x 4 + 2 = 8.67 (about 18% slower) NAND4: d = 2 x 4 + 4 = 12.0 NOR4: d = 3 x 4 + 4 = 16.0 (about 33% slower)
Multiply by the process's basic delay unit (tau) to get time. Whatever the process, the NOR gate is the slower one, and the penalty grows with the number of inputs.
What designers do with this
- Wide NOR functions are rewritten with De Morgan's law (A + B + C + D)' = A' B' C' D', or split into a tree of smaller gates.
- Where speed matters, designers place the latest-arriving input on the transistor closest to the output, which is easier to exploit in a NAND stack.
Interview angle
If asked "why NAND over NOR?", a strong answer mentions the series pMOS stack in a NOR gate, the mobility difference between holes and electrons, and quotes the logical effort of 4/3 versus 5/3 for 2-input gates. Avoid saying NOR needs more transistors; at the same input count both use 2n transistors.


Your comments will be moderated before it appears here.